An experimental study of tremolite dissolution rates as a function of pH and temperature: Implications for tremolite toxicity and its use in carbon storage

Steady-state tremolite dissolution rates, at far-from-equilibrium conditions, were measured as a function of aqueous silica and magnesium activity, pH from 1.9 to 6.7, and temperature from 25 to 150ºC. Calcium is released from tremolite faster than either Mg or Si throughout most of the experiments even after these latter elements attained steady-state release rates. The preferential removal of Ca releases fine Mg-Si rich needle-like fibres from the tremolite, probably promoting its toxicity. In contrast, Mg was released in stoichiometric or near to stoichiometric proportion to Si once steady-state was attained. Measured steady-state tremolite dissolution rates based on Si release can be described using where r+ signifies the BET surface area-normalized forward tremolite steady-state dissolution rate, AA refers to a pre-exponential factor = 6610–3 mol cm–2 s–1, EA designates an activation energy equal to 80 kJ mol–1, R represents the gas constant, T denotes absolute temperature, and ai refers to the activity of the subscripted aqueous species. This rate expression is consistent with tremolite dissolution rates at acidic pH being controlled by the detachment of partially liberated silica tetrahedra formed from the exchange of Mg2+ for two protons near the mineral surface after the near-surface Ca has been removed. Nevertheless, Mg release rates from tremolite are ~3 orders of magnitude slower than those from forsterite and enstatite suggesting that tremolite carbonation will be far less efficient than the carbonation of these other Mg-silicate minerals.

The motivation for this study is several fold: (1) The dissolution rates of Ca-and Mg-bearing silicates, such as tremolite, are of particular interest as they are potential sources of the divalent metals required for the mineralogical storage of CO 2 (cf. Oelkers and Schott, 2005;Hänchen et al., 2008;Oelkers and Cole, 2008;Oelkers et al., 2008a;King et al., 2010;Ryu et al., 2011;Saldi et al., 2012;Gislason and Oelkers, 2014).
(3) The dissolution rates of the Mg silicates appear to exhibit a systematic behaviour as a function of aqueous solution pH and composition . It is anticipated that this study will provide further insight into this systematic behavior of these minerals.
The goal of this study is to expand our knowledge of tremolite dissolution kinetics at acid to neutral pH. Thus tremolite dissolution rates have been measured at 25 to 150ºC, and 1.9 < pH < 6.7 as a function of aqueous fluid concentration at far-from-equilibrium conditions. The results of this study illuminate the durability of tremolite at conditions relevant to mineral carbonation and human health.
To date there have been relatively few studies of the dissolution rates of tremolite reported in the literature. Schott et al. (1981) reported the dissolution rates of tremolite at 22ºC at 1 4 pH 4 6 in aqueous HCl solutions measured in batch reactors. Mast and Drever (1987) reported the dissolution rates of tremolite at 22ºC and 2 4 pH 4 9 in aqueous solutions containing HCl, KHCO 3 , H 3 BO 3 and KOH measured in fluidized bed reactors. Measured rates decrease slightly with increasing pH to at least pH 9. In addition, both of these experimental studies observed a preferential release of Mg over Si during the early stages of experiments performed at acidic conditions prior to attainment of stoichiometric steady-state dissolution. This behaviour and similar observations on other divalent metal silicates reported by Oelkers et al. (2009) were attributed to a relatively rapid exchange reaction between H + and Mg 2+ on the mineral surfaces. Similarly, Schott et al. (1981) reported that Ca was preferentially released compared to both Mg and Si during tremolite dissolution at acid conditions. Mast and Drever (1987) performed additional tremolite dissolution experiments at pH 4 and 7 as a function of aqueous oxalate concentrations. In contrast to the effect of the presence of aqueous organic species on Al-silicate minerals (cf. Oelkers and Schott, 1998), no effect of the presence of aqueous oxalate was observed for oxalate concentrations 410 À3 M. The present study expands upon this past work by determining tremolite steady-state dissolution rates over a wide range of pH at temperatures between 25 and 150ºC and by establishing an equation to describe these rates over a variety of environmental conditions.

Theoretical background
The standard state adopted in this study is that of unit activity for pure minerals and H 2 O at any temperature and pressure. For aqueous species other than H 2 O, the standard state is unit activity of the species in a hypothetical 1 molal solution referenced to infinite dilution at any temperature and pressure. Tremolite dissolution can be described according to: Note that SiO 2 (aq) in reaction 1 represents the H 4 SiO 4 neutral aqueous species rather than total dissolved Si. Taking account of the standard state, the law of mass action for reaction 1 can be written where K tremolite stands for the equilibrium constant of reaction 1, and a i represents the activity of the subscripted aqueous species. The chemical affinity (A) for reaction 1 can be expressed as where R designates the gas constant, and T represents absolute temperature. When the fluid is undersaturated with respect to tremolite, A is negative, and when the fluid is supersaturated, A is positive. All thermodynamic calculations reported in the present study were performed using the PHREEQC 2.6 computer code (Parkhurst and Appelo, 1999) together with its llnl database (Johnson et al., 2000). It is assumed in all thermodynamic calculations that the solid used in the dissolution experiments was pure stoichiometric tremolite. Although surface complexation theory suggests that the activities of charged particles on charged surfaces may differ somewhat from unity (Davis and Kent, 1990), the activities of species at the tremolite surface are assumed to be equal to their mole fraction. This latter assumption is shown below to be consistent with measured rate data. Within the context of Transition State Theory, surface reaction-controlled dissolution rates can be considered to be the difference between the forward rate (r + ) and the reverse rate (r À ) such that Taking account of the law of detailed balancing it can be shown that equation 4 is equivalent to Helgeson, 1977, 1982;Lasaga, 1981;Schott and Oelkers, 1995) where s stands for Temkin's average stoichiometric number equal to the ratio of the rate of destruction of the activated or precursor complex relative to the overall dissolution rate. The form of equation 5 is such that overall rates (r) equal forward rates (r + ) when A >> sRT. All dissolution rates in the present study were measured at farfrom-equilibrium conditions, such that A >> sRT. At these conditions r À << r + and thus r & r + . Tremolite dissolution rates in this study are thus symbolized r + . Such experimental results can be used to assess the effect of aqueous solution composition on forward dissolution rates independently of the effects of chemical affinity.
Steady-state forward tremolite dissolution rates in this study are quantified using the multi-oxide silicate dissolution model of Oelkers (2001a). Within this model, dissolution proceeds via the sequential breaking of metalÀoxygen bonds until the mineral structure is destroyed. The tremolite structure consists of CaÀO and MgÀO bonds in 8-fold and 6-fold coordination, respectively, and tetrahedral SiÀO bonds (Hawthorne and Oberti, 2007). Silicon tetrahedra form double chains layered between strips of edge-sharing octahedra. In accord with Oelkers (2001a) and consistent with the results reported below, the relative rate of breaking these bonds are CaÀO > MgÀO > SiÀO. This conclusion is consistent with the observations that Mg and Ca are initially released more rapidly than Si from divalent metal silicates (e.g. Luce et al., 1972;Lin and Clemency, 1981;Schott et al., 1981;Berner and Schott, 1982;Petit et al., 1987;Oelkers and Schott, 2001;. Such observations suggest that the tremolite dissolution mechanism is initiated by the relatively rapid removal of Ca atoms and Mg atoms via Ca-H and Mg-H exchange reactions (Oelkers, 2001a). This exchange reaction partially liberates Si-O tetrahedra, which are subsequently released to aqueous solution, completing the dissolution process. In accord with transition state theory, far-from-equilibrium tremolite dissolution rates will therefore be proportional to the concentration of partially detached Si-tetrahedra (cf. Oelkers 2001a; Schott et al., 2009) such that In equation 6 [> Si*] represents the concentration of partially detached Si tetrahedra at the tremolite surface, k + designates a rate constant, a i again refers to the activity of the subscripted aqueous species, M i and z i stand for the identity and the charge of the cations exchanged with protons to create the partially liberated Si tetrahedra (in the case of tremolite, M z þ i i refers to Ca 2+ and Mg 2+ ), n denotes a stoichiometric coefficient equal to the number of partially detached Si tetrahedra formed by the removal of each divalent cation, and K i designates the equilibrium constant for the metalproton exchange reactions. Note that the product on the right hand side of equation 6 calculates the concentration of partially detached Si tetrahedra forming due to protonÀmetal exchange reactions. In cases where all of one type of cation is removed from the mineral surface, the parenthetical term on the right side of equation (6) reduces to unity; in cases where there are relatively few of one type of cations removed from the surface, this parenthetical term reduces to (see Oelkers and Schott, 2001): In the case of tremolite at the conditions considered in the present study, Ca is removed rapidly from the mineral surface and continues to be removed~6 times more rapidly, when normalized for its stoichiometry, than either Mg or Si even after these latter two elements attain a steady-state release rate. Magnesium, in contrast, is released in close-to-stoichiometric quantities compared to Si at steady-state (see below). In such cases equation 6 reduces to (see Oelkers, 2001a;Oelkers and Schott, 2001;Saldi et al., 2007) where k + ' refers to the product k + K Mg . This equation suggests that increasing aqueous hydrogen ion activity and decreasing aqueous Mg activity will increase far-from-equilibrium tremolite dissolution rates and these rates will be independent of aqueous Si activity.
The variation in rate constants with temperature is commonly described using an Arrhenius equation such as (Arrhenius, 1889): where A A stands for a pre-exponential factor, E A is the activation energy, R refers to the gas constant and T signifies the absolute temperature. Combining equations 7 and 8 gives an equation describing tremolite forward dissolution rates with respect to the temperature and aqueous solution composition: The degree to which equation 9 provides a quantitative description of tremolite dissolution rates is assessed below.

Methods and samples
Two different tremolite samples were used in this study, providing some variation in tremolite morphology in the starting material. One sample (tremolite A) was a massive sample of bladed to prismatic tremolite crystals from Felch, Michigan, USA, and the other (Tremolite B) was a fibrous tremolite sample from Cannan, Connecticut, USA. Both samples were crushed using a pulverizer and agate mortar and pestle. Selected size fractions of this powder were then cleaned ultrasonically using acetone. The resulting tremolite powders were Two distinct open-system reactor systems were used in this study. Experiments were performed at 25, 37 and 50ºC in 250 ml Azalon plastic mixedflow reactors placed in a thermostated bath (see Chairat et al., 2007). These reactors were fitted with Nalgene tubes for supply and recovery of experimental fluids, which were pumped through the reactor with Gilson peristaltic pumps. The reacted fluids were passed through a 0.45 mm Millipore Nitrocellulose in-line filter upon exiting the reactor. Floating teflon-coated stir bars were used to mix the powder/fluid mixture in the reactor and to avoid grinding. Experiments were performed at 80, 100 and 150ºC in a Parr titanium mixed-flow reactor system, such as described by Oelkers et al. (2008b). A High Precision/High Pressure Liquid Chromatography Pump provided continuous fluid flow ranging from 1 to 2 g/min during the experiments. The precision of the fluid flow rates was Ô4%. The volume of the titanium reactor was 250 ml. The fluid within the reactor was stirred using a Parr magnetically driven stirrer, the temperature controlled by a Parr controlled furnace, and elevated pressure was maintained using a back-pressure regulator.
Experiments were performed in this study either individually or in series, consisting of several experiments performed on a single tremolite powder. Each experimental series is distinguished by the first two entries of its experiment number; the first entry indicates the identity of the tremolite powder used, and the second letter the experimental series run with this powder. For example, series A-a, was performed using tremolite A and consists of experiments A-a-1, A-a-2, and A-a-3. Note that some series consisted of just one experiment. At the beginning of each experimental series the reactor was dismantled at ambient conditions. A specific mass of tremolite powder was placed in the reactor. The reactor was filled with the initial inlet fluid, closed, and placed in a thermostated bath or in a furnace. The temperature, pressure, fluid flow, and stirring rate were adjusted to desired settings. The fluid flow rate and outlet fluid composition were measured regularly. After steady state was verified with a minimum of three constant Si concentrations in the outlet fluid samples obtained over several residence times (defined as the volume of the reactor divided by the reactive fluid flow rate): either (1) the experimental series was stopped; or (2) the inlet fluid composition, fluid flow rate and/or the temperature were then changed to the next desired experimental condition.
Dissolution experiments were performed in fluids consisting of demineralized H 2 O, and reagent-grade HCl, NaCl, CaCl 2 and MgCl 2 from Merck, and aqueous SiO 2 created from the dissolution of amorphous SiO 2 at 80ºC. All inlet fluids had stoichiometric ionic strengths of 0.02 or 0.12 mol kg À1 . The compositions of all inlet fluids are listed in Table 2. Calcium and magnesium compositions of the inlet and outlet fluids were determined using a Perkin Elmer Zeeman 5000 atomic absorption spectrometer. Silica compositions were measured using the Molybdate Blue method (Koroleff, 1976). The reproducibility of chemical analyses was Ô4% for Si, Mg and Ca concentrations >0.5, >0.1 and >0.1 ppm, respectively, but as much as Ô10% at lower concentrations. Outlet fluid pH was measured at 25ºC immediately after sampling. All outlet fluids were undersaturated with respect to all possible secondary phases as confirmed by PHREEQC calculations.
Steady-state forward dissolution rates (r + ) were computed from the measured steady-state solution compositions using where Dm i stands for the concentration difference between the inlet and outlet of the ith element in solution, F represents the fluid mass flow rate, n i refers to the stoichiometric number of moles of the ith element in one mole of tremolite, and s BET denotes the BET specific surface area of the initial tremolite, and M signifies the initial mass of tremolite in the reactor.

Results
Photomicrographs of the tremolite surfaces following their dissolution during two representative experiments are shown in Fig. 1c,d. Two observations are apparent. First, dissolution appears to occur primarily on the tremolite prism edges, where the divalent cations can readily move out of the structure. Secondly, a significant number of fine-grained needle-shaped particles are present on the surfaces following the experiments. Thirty four steady-state dissolution rates were obtained in this study; experimental results and conditions are summarized in Table 3. Two examples of the temporal evolution of instantaneous rates generated from Si, Mg and Ca release are shown in Fig. 2. The first example illustrates    rates during experimental series B2-g, performed at pH 4.3 and temperatures of 37 and 50ºC. The initial release rates of both Mg and Ca are approximately an order of magnitude faster than that of Si after normalization to the stoichiometry of the dissolving tremolite. The release rates of Mg decrease with time to be stoichiometric with respect to Si after~25 days of elapsed time, and remain near to stoichiometric with respect to Si until the end of the experimental series. In contrast, the Ca release rates remain~0.6 orders of magnitude faster until the end of this series. The second example illustrates the instantaneous tremolite dissolution rates in experiment B2-e-1 performed at pH 6.9 at a temperature of 37ºC. Initially, Mg and Si are released in stoichiometric Log of the saturation index of the steady-state outlet fluid with respect to tremolite as calculated using PHREEQC.

EXPERIMENTAL STUDY OF TREMOLITE DISSOLUTION RATES
proportions, with Ca being released more than an order of magnitude faster. As Si release rates approach steady state, the Mg release rate appears to increase and its release rates are~0.3 orders of magnitude greater than that of Si at the end of the experiment.
The logarithms of steady-state tremolite dissolution rates based on Mg and Ca release are compared with those based on Si release in Fig. 3. Note that the definition of steady state adopted in this study is that of a near constant Si release rate; Mg and Ca may not have attained a steady-state in these experiments. Nevertheless, as can be seen in Fig. 3, tremolite dissolution rates based on Mg release generally fall within 0.3 log units of the corresponding Si rates. In contrast, steady-state rates based on Ca release are, on average, six times greater than corresponding rates based on Si release. Note that rates based on Ca release measured at the elevated temperature are closer to or even lower than their corresponding Si rates determined at temperatures >80ºC The variation of forward steady-state tremolite dissolution rates measured at 37 and 50ºC based on Si release are depicted as a function of pH in Fig. 4 and as a function of logða 2 H þ =a Mg 2þ Þ in Fig. 5. The general trend of the rates is to decrease with increasing pH with a log rate vs. pH slope near À0.25. Similarly, despite scatter, rates tend to increase with increasing logða 2 H þ =a Mg 2þ Þ; the dashed lines drawn in Fig. 5, consistent with the symbols, have a slope of 0.125, consistent with n = 8 in equation 7.
The variation of measured steady-state tremolite dissolution rates at pH~2 as a function of temperature are quantified in the present study using the empirical Arrhenius equation, equation 8. The activation energy and preexponential constant in equation 9 were obtained with the aid of the plot shown in Fig. 6. Regression of these data, as illustrated in this figure, yields E A = 80 kJ/mol. The distribution of rates and the curves shown in Figs 5 and 6 suggest that tremolite steady-state dissolution rates can be described as a function of temperature and fluid composition using The degree to which equation 11 describes the rate obtained in the present study can be assessed with Fig. 7 which compares rates determined using equation 11 with their measured counterparts. The average difference between calculated and measured rates is 0.3 log units. Comparison of all forward steady-state tremolite dissolution rates measured in this study with those calculated using equation 11. The symbols represent measured and calculated rates, whereas the dashed line corresponds to equal values of these two rates. The error bars are consistent with a Ô0.15 uncertainty in the rates.

Calcium release and tremolite toxicity
The preferential release of Ca has been observed throughout all experiments performed at T<80ºC. Similar to the behaviour of wollasonite during its dissolution ), Ca appears to be removed more quickly than and independently of the other elements. In the case of wollastonite, the Ca-depleted layer does not attain a steady-state thickness at acidic conditions, but grows continuously. Such also appears to be the case for tremolite at least for the duration of the <80ºC experiments; the tremolite structure has channels parallel to its Si tetrahedral chains allowing for the ready removal of Ca through proton-exchange reactions. A significant observation is that the tremolite structure consists of linear talc-like Mg-Si ribbons, with the Mg in 6-fold coordination bound together by adjoining Ca in 8-fold coordination (Warren, 1930;Hawthorne and Oberti, 2007). A schematic illustration of the tremolite structure, projected down its c axis is shown in Fig. 8. It is clear from this image that the removal of Ca will liberate the talc-like ribbons from the bulk tremolite, creating fine needle-like particles. Such particles are evident from the tremolite recovered following the dissolution experiments performed in the present study. It follows that the preferential Ca release in lung fluids could promote the toxicity of inhaled tremolite through the creation of such fine-grained needle-like particles.

Comparison of the dissolution rates of Mg-silicate minerals
The dissolution rates of selected Mg-silicate minerals at ambient temperatures are illustrated in Fig. 9. The rates shown were selected because they were all generated using similar techniques in the same laboratory, facilitating their comparison. Mg-silicate dissolution rates all tend to decrease with increasing pH, though there is some indication that these rates become pH independent at alkaline conditions. Notably, the dissolution rates decrease substantially with the degree of connectiveness of the Si structure. Forsterite, consisting of isolated Si tetrehedra, dissolves faster than enstatite, which is made up of single chains of Si tetrehedra, which itself dissolves faster than talc and tremolite, consisting of sheets of Si tetrehadra, and talc-like double-chained Si tetrehedra. The similar talc and tremolite dissolu- FIG. 9. Comparison of 25ºC forward steady-state dissolution rates as a function of pH for selected Mg-silicate minerals (after Schott et al., 2009) and an aqueous Mg 2+ activity of 1610 À4 . Rates for talc and enstatite were calculated using equations reported by Saldi et al. (2007) and Oelkers and Schott (2001), whereas those for forsterite were reported by Pokrovsky and Schott (2000). Rates for tremolite were calculated using equation 11. tion rates probably reflect the similarity in their Si-Mg-O framework after the Ca is removed from the latter mineral. Taken together, the comparison shown in Fig. 9 is consistent with a similar dissolution mechanism for these minerals comprising the relatively fast removal of Mg followed by the slower rate-limiting liberation of partially detached Si from the mineral surface.

Experimental uncertainties and data scatter
Uncertainties associated with the rates generated in this study arise from a variety of sources, including the measurement of aqueous solution concentrations, fluid-flow rates, and tremolite surface areas. The uncertainties in the measured values of total aqueous silica, magnesium, and calcium concentration are on the order of Ô10% or less. Computational and experimental uncertainties in the pH measured are on the order of Ô0.04 pH units. Uncertainties in fluid-flow rate measurements are not more than 4%. The uncertainty associated with the measurement of the surface area of the initial tremolite powder is Ô10%. If uncertainties were estimated exclusively from the sum of these contributions, an overall uncertainty of the dissolution rates reported would be of the order of 20%. Nevertheless it is clear from Figs 4À6 that other sources lead to the additional scatter in the measured rates. According to mass-balance calculations, <5% of the initial tremolite powder was dissolved in any experimental series. It seems possible, however, that the tremolite surface area present in the reactor evolves significantly during the experiments. Although post-experiment surface areas were not measured, as emphasized above, a significant number of fine, needle-shaped gains were formed during the dissolution experiments performed in this study. As emphasized by a number of studies (Gautier et al., 1994;Luttge et al., 2013;Fischer et al., 2014), the degree to which reactive surface area varies in response to dissolution is currently impossible to define unambiguously, yet consideration of the scatter in the Figs 4À6 suggests that such effects may be large. This conclusion is consistent with the rates obtained from experimental series A-a, which dissolved tremolite at pH 2 and 25, 37 and 50ºC. This experimental series leads to an anomalously fast rate during the third experiment performed in the series, after it had already dissolved at pH 2 at 25 and 37ºC, potentially due to the creation of fine particles related to the preferential release of Ca.

Conclusions
The results and discussion presented above illustrate the distinct dissolution behaviour of tremolite. Tremolite dissolution rates based on Si or Mg release are orders of magnitude slower than corresponding dissolution rates of forsterite and enstatite. As a consequence, the carbonation of tremolite will be far less efficient than the carbonation of these other Mg silicate minerals, limiting its potential in mineral carbon storage efforts. The dissolution of tremolite, however, releases calcium preferentially, at least at temperature >80ºC. This preferential release continues for a substantial amount of time, leading to the extensive removal of Ca from the tremolite surface. Due to its mineral structure, the removal of Ca will promote the breaking apart of the tremolite into small needle-shaped particles, which probably will increase its toxicity to human lungs. It seems, therefore, that an improved understanding of the chemical processes controlling the retention vs. the removal of Ca may help to design methods to limit the toxicity of this mineral.