Precipitation of solid phase calcium carbonates and their effect on application of seawater S ATP models

At the present time, little is known about how broad salinity and temperature ranges are for seawater thermodynamic models that are functions of absolute salinity (SA), temperature ( T ) and pressure ( P ). Such models rely on fixed compositional ratios of the major components (e.g., Na/Cl, Mg/Cl, Ca/Cl, SO 4/Cl, etc.). As seawater evaporates or freezes, solid phases [e.g., CaCO 3(s) or CaSO42H2O(s)] will eventually precipitate. This will change the compositional ratios, and these salinity models will no longer be applicable. A future complicating factor is the lowering of seawater pH as the atmospheric partial pressures of CO2 increase. A geochemical model (FREZCHEM) was used to quantify theSA-T boundaries atP=0.1 MPa and the range of these boundaries for future atmospheric CO2 increases. An omega supersaturation model for CaCO 3 minerals based on pseudo-homogeneous nucleation was extended from 25–40 C to 3C. CaCO3 minerals were the boundary defining minerals (first to precipitate) between 3C (at SA=104 g kg−1) and 40C (at SA=66 g kg−1). At 2.82C, calcite(CaCO3) transitioned to ikaite(CaCO 36H2O) as the dominant boundary defining mineral for colder temperatures, which culminated in a low temperature boundary of−4.93C. Increasing atmospheric CO 2 from 385μatm (390 MPa) (in Year 2008) to 550 μatm (557 MPa) (in Year 2100) would increase the SA and t boundaries as much as 11 g kg−1 and 0.66C, respectively. The model-calculated calcite-ikaite transition temperature of 2.82 C is in excellent agreement with ikaite formation in natural environments that occurs at temperatures of 3 C or lower. Furthermore, these results provide a quantitative theoretical explanation Correspondence to: G. M. Marion (giles.marion@dri.edu) (FREZCHEM model calculation) for why ikaite is the solid phase CaCO3 mineral that precipitates during seawater freezing.


Introduction
At the present time, little is known about how broad salinity and temperature ranges are for seawater thermodynamic models that are functions of absolute salinity (S A ) [or practical salinity(S)], temperature (T ) and pressure (P ), such as Feistel (2003Feistel ( , 2008) ) and Feistel and Marion (2007).Models that are used rely on fixed chemical composition ratios (Na/Cl, Ca/Mg, Cl/HCO 3 , etc., Millero et al., 2008).As seawater evaporates or freezes, eventually solid phases such as aragonite (CaCO 3 ), calcite (CaCO 3 ), or ikaite (CaCO 3 6H 2 O) will precipitate.When precipitation happens, the fixed ratios of various soluble species will change and the seawater salinity (S A ) model will no longer be applicable.
Aragonite and calcite are generally the first solid phase salts that precipitate during seawater evaporation (Pytkowicz, 1973;McCaffrey et al., 1987;Maldonado et al., 1992;Morse et al., 1997;Zuddus and Mucci, 1998;Hardie, 2003;Millero, 2007;Morse et al., 2007).Aragonite and calcite have identical chemical compositions, but differ in their crystal structures and solubilities.How to handle these carbonates with chemical thermodynamic models is problematic because surface seawater is always supersaturated with these minerals.Fortunately, Morse and He (1993) have experimental data that calculates the degree of supersaturation ( ) at the point where CaCO 3 minerals will start to precipitate from seawater.Unfortunately, these datasets are only defined for 25 and 40 • C. A controversy that dates back a century deals with whether calcite or ikaite precipitates during seawater freezing (Ringer, 1906;Gitterman, 1937;Assur, 1958;Richardson, 1976;Weeks and Ackley, 1982;Marion, 2001;Dieckmann et al., 2008).Recently, for the first time, experimental measurements in Antarctic sea ice discovered ikaite crystals had formed during seawater freezing (Dieckmann et al., 2008).But how supersaturated calcite and saturated ikaite can be integrated into a theoretical model still remains an open question.
Another complicating factor for the long-term application of ranges for S A -T -P models is the global rise of atmospheric CO 2 partial pressure from a pre-industrial (1750-1800) level of 280 µatm, to 385 µatm in 2008, and potentially to 550 µatm in 2100 (http://www.esrl.noaa.gov/gmd/ccgg/trends/).Such CO 2 trends will affect seawater pH and carbonate mineral solubilities that will alter S A -T -P applicability boundaries.
The specific objectives of this paper were to (1) establish salinity-temperature boundaries for S A -T -P models at P =0.1 MPa (Earth surface pressure), (2) establish the range of these boundaries for future atmospheric CO 2 increases, and (3) reconcile, if possible, the controversy dealing with calcite-ikaite equilibrium during seawater freezing.
The S A -T boundaries for seawater S A -T -P models will be established with a theoretical chemical thermodynamic model called FREZCHEM (Marion and Kargel, 2008).The composition of seawater will be based on a newly established standard composition based on S A (Millero et al., 2008).And finally, CaCO 3 supersaturation at precipitation will be integrated into FREZCHEM based on a pseudo-homogeneous nucleation model (Morse and He, 1993).
This Pitzer-based model was used in Feistel and Marion (2007) to extend an S A -T -P model (Feistel, 2003) from S A =42 g kg −1 to 110 g kg −1 and the cold temperature boundary from −2 to −7 • C.This model contains calcite, aragonite, vaterite, and ikaite chemistries.What FREZCHEM currently lacks in addressing the objectives of this paper is how to cope with CaCO 3 supersaturation in seawater, which we will address below.

Seawater properties
The seawater molalities [moles kg(H 2 O) −1 , m] that will drive our simulations are taken from Millero et al. (2008) (Table 1).What is classified as alkalinity in Table 1 is "fixed" through charge balance with the other listed constituents.These tabular values will only change when S A changes.As T and/or P CO 2 change, these tabular values can cause minor changes on S A , unless solid phases start to precipitate.The minor effect on changing S A during T and P CO 2 change is estimated to be about 0.04%, based on soluble carbon contents, which can be safely ignored.Whether we represent the S A values as 35.147 g/kg or as 35.133 g/kg is minor.On the other hand, as T and/or P CO 2 change, the individual alkalinity constituents (Table 1 footnote) can make major changes due to the effects of P CO 2 and T on controlling pH.For example at S A =35.147 g/kg with P CO 2 =385 µatm, the pH at 40 • C is equal to 8.318 with CO 2− 3 =1.21e-4m, while pH at 0 • C is equal to 8.056 with CO 2− 3 =4.04e-5m.These major changes are factored into model calculations.In addition to these ions, we also included B(OH) 3 =0.00033m.Several minor constituents (Sr 2+ , B(OH) − 4 , F − ) were excluded from our simulations.Overall, there was a small decrease in the standard S A from 35.165 g kg −1 (Practical Salinity S=35.00) to the model-used 35.147 g kg −1 .Our research efforts will focus exclusively on Absolute Salinity (S A ).To convert S A to S (Practical Salinity on the 1978 scale), one can use the relationship: S=(35/(35.16504g kg −1 )) S A =0.995307 S A /(g kg −1 ) (Millero et al., 2008).Throughout the text, we will represent atmospheric CO 2 in units of µatm (=MPa/1.01325)as P CO 2 .But the FREZCHEM model actually uses fugacity (f CO 2 ) in model calculations.

CaCO 3 nucleation/supersaturation
There are distinct differences among homogeneous, pseudohomogeneous, and heterogeneous nucleation and precipitation of CaCO 3 minerals.Homogeneous precipitation of CaCO 3 occurs in the absence of pre-existing precipitation surfaces (solid or soluble phases).Pseudo-homogeneous precipitation occurs on soluble colloidal particles, microbes, planktonic debris, and POC or reaction vessel surfaces (Morse and He, 1993).Heterogeneous precipitation occurs on pre-existing mineral solid phases.Heterogeneous precipitation is relevant where carbonate solid phases are in close contact with supersaturated waters (e.g., in shallow waters or in the presence of biotic carbonates).We will focus our attention in this paper primarily on pseudo-homogeneous CaCO 3 precipitation using the Morse and He (1993) experimental data developed from natural seawater that likely contained surfaces (soluble and reaction vessel) for CaCO 3 precipitation.We will also discuss two examples of heterogeneous precipitation that can be important in the interpretation of experiments with added crystals, and seawater environments with natural crystals.
In the seawater literature, it has long been known that seawater is supersaturated with carbonate minerals such as calcite and aragonite (Morse and Mackenzie, 1990;Millero and Sohn, 1992;Millero, 2007;Morse et al., 2007).Traditionally, the degree of supersaturation was described by the omega ( ) concept: where IAP is the calculated ion activity product [e.g., (a Ca )(a CO 3 )] for a specific mineral (e.g., calcite), and K is the solubility product for the mineral.For example, for calcite at S A =35.147 g kg −1 and t=25 • C is calculated to be 5.6 with the FREZCHEM model (IAP=1.856e-8,K=3.312e-9).The seawater sample is 5.6-fold supersaturated with respect to calcite.While this is interesting, it provides no clue as to when such a seawater sample would actually start to precipitate a carbonate mineral during the evaporation or freezing process as S A increases.
John Morse and colleagues (Morse and He, 1993;Morse et al., 2007) have developed an empirical omega model that is geared to estimating the point where carbonate minerals will actually start to precipitate as a function of S and T .In this case, the omega concept is defined as: In Eq. ( 2), the IAP is based on experimental measurements and theoretical calculations at the point where a "CaCO 3 " mineral in natural seawater started to precipitate, likely either as calcite or aragonite with Mg substitutions.The K(calcite) is a reference point based on the equilibrium of pure calcite.Despite the fact that Eq. ( 2) is defined in terms of calcite, the IAP(CaCO 3 ) only predicts that an undefined CaCO 3 mineral will precipitate.In what follows, we will use: to calculate the IAP(CaCO 3 ) at the point where a CaCO 3 mineral will start to precipitate.We could also have used an aragonite reference state [K(aragonite)].But in either case, the IAP(CaCO 3 ) is identical (Eq.3).Effectively, IAP(CaCO 3 ) is the equilibrium constant for CaCO 3 minerals in seawater.We will primarily run the FREZCHEM model at a fixed T under the evaporation option, where S A is steadily increasing because water is being removed, until our modelcalculated IAP agrees with the empirical IAP (CaCO 3 equilibrium constant) of Eq. ( 3) (or another mineral).At this point, the calculated salinity (S A ) represents the boundary for our S A -T -P model because a CaCO 3 (or another) mineral has started to precipitate.We will also run the FREZCHEM model under the freezing option to establish the lower temperature boundary for the S A -T -P model.Bear in mind that while Eq. ( 3) specifically addresses CaCO 3 mineral formation, this concept integrated into the FREZCHEM model does not preclude the possibility that other minerals (with other ions) could precipitate first during the evaporation or freezing process.Where solutions have become supersaturated with respect to specific minerals, the FREZCHEM model selects the solid phase that minimizes the Ca (or other ion) concentration as the most thermodynamically stable mineral."CaCO 3 " as defined in Eq. ( 3) is the only solid phase in the FREZCHEM model that is allowed to persist at the supersaturated level before precipitating.On the other hand, coping with the degree of supersaturation associated with Eq. ( 3) required that we remove several carbonate minerals from the FREZCHEM mineral database, including aragonite (CaCO 3 ), vaterite (CaCO 3 ), dolomite [CaMg(CO 3 ) 2 ], magnesite (MgCO 3 ), and hydromagnesite [3MgCO 3 Mg(OH) 2 3H 2 O].These minerals, except for aragonite that is a factor in Eq. (3), do not precipitate from seawater despite their supersaturation.So, removing them from the mineral database is perfectly valid for seawater simulations.Retention of the latter carbonate minerals in FREZCHEM would lead to their precipitation rather than the supersaturated "CaCO 3 " as defined in Eq. (3).As pointed out above, "CaCO 3 " likely represents either aragonite or calcite.

Results
In Fig. 1  nucleation (Morse and He, 1993;Morse et al., 2007).The S values of the original data were converted to S A values (Fig. 1) by S A = 1.004715 (g kg −1 ) S. We fit a linear equation to these data, which is given by: where r 2 (coefficient of determination)=0.922, and SE (standard error)=0.54(Steel et al., 1997).The three coefficients in Eq. ( 4) (and Fig. 1) are all statistically significant at the <1% chance of a Type I error (Steel et al., 1997).Extrapolation of the equation to 0 • C is also depicted in Fig. 1.The equation fits to data at 25 and 40 • C seem reasonably good.The extrapolation to lower temperatures will be examined below (see Sect. 4 on Discussion).Now given an equation that allows one to calculate (calcite) as a function of S A and T (Eq.4), we can estimate the IAP(CaCO 3 ) of Eq. ( 3).Next, we ran the FREZCHEM model under the evaporation process, at a fixed T and a starting S A =35.147 g kg −1 , until the model-calculated IAP equilibrated with IAP(CaCO 3 ) of Eq. (3) to form carbonates, or other minerals.This process allowed us to place S A and T limits where solid phase minerals start to form.The results of these calculations for Year 2008 with CO 2 =385 µatm are given in Fig. 2 (solid lines).As an example, for a temperature of 20 • C with initial S A =35.147 g kg −1 , we evaporated the solution with FREZCHEM until CaCO 3 started to precipitate at S A =83.1 g kg −1 , which sets the boundary limit for an (Fig. 2).Aragonite is likely the dominant CaCO 3 mineral between 8 and 40 • C, and calcite is likely the dominant CaCO 3 mineral between 3 and 8 • C (Morse et al., 2007).Note that in this temperature range of 3-40 • C, the boundary increases with decreasing temperature because the solubility of CaCO 3 minerals increases with decreasing temperature.At a temperature of 2.82 • C, the FREZCHEM model predicts that ikaite should start precipitating instead of CaCO 3 .Because ikaite solubility decreases with decreasing temperature, the slope of the curve changes from that for CaCO 3 solubility (Fig. 2).Now to establish the lower temperature limit, we needed to extend the calculations to the point where minerals start precipitating at cold temperatures.If we start at 25 • C with S A =35.147 g kg −1 , then lowering temperature will eventually reach the "ice line" in Fig. 2 at t=−1.9 • C. Ice formation, per se, does not limit the range of S A -T -P models because relatively pure ice forms concentrating all solution phase constituents equally, therefore leaving ratios such as Ca/Mg unchanged.But as temperature continues to changes below −1.9 • C, the freezing process will follow the ice line changing rapidly in S A until it hits the ikaite line at −4.93 • C with S A =87.3 g kg −1 (Fig. 2).The lower t limit for all phases that lead to ice formation is −4.93 • C.
The atmospheric CO 2 has changed historically from 280 µatm in pre-industrial years, to 385 µatm in 2008, and potentially to 550 µatm by 2100 (http://www.esrl.noaa.gov/gmd/ccgg/trends/).The corresponding increases in P CO 2 in surface ocean water results in a decrease in seawater pH (at S A =35 g kg −1 , t=25 • C) from 8.38 in pre-industrial years, to 8.27 in 2008, and to 8.14 in 2100 (Fig. 3).These pH values were estimated with the FREZCHEM model that defines pH=−log 10 (H + ), where the parentheses define H + activity.In contrast, had we estimated pH Ocean Sci.  using the seawater scale (SWS) (Millero, 2001) where the brackets define concentrations, then the corresponding pH values are 8.13, 8.02, and 7.88, respectively, about 0.25 pH units lower than the pH values in Fig. 3. Either scale can be used to define pH in seawater.But, caution must be used to assure consistency in how the pH options are defined and subsequent applications to H + equilibria.See Millero (2001) for more details on pH scale variability.
Changing S A , especially lower values, will have a major influence on the relationship of CO 2 and pH (Fig. 3).The Baltic Sea is a classic case where seawater S A is highly variable, from below 1 g kg −1 in the coastal lagoons, over 8-12 g kg −1 in the central Gotland Sea, to about 25 g kg −1 at the Danish Straits, with significant temporal variability from the daily to the decadal (climatological) time scales (see e.g.online figures of Feistel et al., 2008).In addition, brackish Baltic seawater contains anomalously high CaCO 3 concentrations (Rohde, 1966;Millero and Kremling, 1976;Feistel and Weinreben, 2008).
A decreasing pH due to increasing atmospheric CO 2 will cause CaCO 3 minerals to increase in solubility, which will increase the salinity-temperature boundaries by up to 11 g kg −1 and 0.66 • C by 2100 (dashed lines, Fig. 2).Also, interestingly, the model predicts that gypsum (CaSO 4 2H 2 O) should be the boundary precipitating mineral between 0 and 6 • C in 2100 (Fig. 2).We also made an approximation of where gypsum should start to first precipitate between 2008 and 2100, which is at S A =109.5 g kg −1 , t=2.95 • C, and CO 2 =450 µatm.

Discussion
Our efforts have resulted in placing S A -T boundaries for using S A -T -P models at P =0.1 MPa, which was the primary purpose of this paper.But Figs. 1 and 2 deal with pseudohomogeneous nucleation of solid phase CaCO 3 .McCaffrey et al. (1987) estimated that CaCO 3 starts precipitating from seawater at 1.8-fold brine concentration at a temperature of ≈31 • C.Those experiments were made at the Morton Bahamas solar salt production facility on Great Inagua Island in the Bahamas and represent heterogeneous nucleation because of the shallow depths that likely contain CaCO 3 solid phases (likely aragonite) nucleation sites.Multiplying 1.8×35 leads to a salinity of 63 g kg −1 at the point where CaCO 3 starts to precipitate.This result compares to a salinity of 73 g kg −1 at 31 • C under pseudo-homogeneous nucleation (Fig. 2).Gitterman (1937) examined CaCO 3 formation resulting from seawater freezing where his experimental aqueous solutions (S A =33.34 g kg −1 ) were likely seeded with calcite crystals (Marion, 2001).His first experimental point, after freezing started at -1.8 • C, was at −3.5 • C where CaCO 3 was clearly precipitating.Because he seeded the aqueous phase with CaCO 3 crystals, this represents heterogeneous nucleation.The brine concentration factor between −1.8 and −3.5 • C is 2.0-fold, which represents a maximum concentration factor because CaCO 3 could have started precipitating at any temperature between −1.8 and −3.5 • C. The point is that heterogeneous nucleation at a temperature that could be as low as −3.5 • C still led to a lower brine concentration than was the case for ikaite that precipitated at a temperature of −4.9 • C under pseudo-homogeneous precipitation (Fig. 2).Heterogeneous nucleation in the presence of CaCO 3 minerals will reduce the applicability range of S A -T -P models.Unfortunately, the data to place a broad range on heterogeneous nucleation is relatively limited.Fortunately, pseudo-homogeneous nucleation (Fig. 2) is more realistic for the bulk of Earth seawater.The experimental data for Figure 1 were based on natural seawater that probably contained soluble nucleation sites associated with clay particles, microbes, planktonic debris, and POC, which is why Morse and He (1993) defined the process as pseudo-homogeneous nucleation of CaCO 3 .This is an appropriate nucleation process for natural seawater, which we modeled in Fig. 2.
What is perhaps most intriguing about the theoretical model calculations is that the model predicts a transition from CaCO 3 (probably calcite) to ikaite at a temperature of 2.82 • C and a salinity of 104 g/kg (Fig. 2).Natural occurrences of ikaite appear to have water temperatures of 3 • C or lower (Pauley, 1963;Bischoff et al., 1993;Larsen, 1994;Omelon et al., 2001), which is in excellent agreement with our model calculation.In the Dieckmann et al. (2008) paper that recently identified ikaite formation in sea ice, great care was taken to keep the temperature of the isolated ikaite crystals between 0 and 2 forming calcite crystals (Omelon et al., 2001;Dieckmann et al., 2008).At the transition temperature of 2.82 • C, the (calcite) value is 25.9, meaning that calcite is 25.9-fold supersaturated.The fact that this high (calcite) value accurately predicts the transition from supersaturated calcite to saturated ikaite at 3 • C is strong support for the broad-range equation for (calcite) (Eq.4, Fig. 1) in which we extended the temperature from 25-40 • C to 3 • C and salinity from 35-70 g/kg (Fig. 1) to 104 g/kg (Fig. 2).Furthermore, this result argues in favor of ikaite formation during the freezing process, which, as we mentioned earlier, has been controversial for over a century (Ringer, 1906;Gitterman, 1937;Assur, 1958;Richardson, 1976;Weeks and Ackley, 1982;Marion, 2001;Dieckmann et al., 2008).The fact that the theoretical FREZCHEM model can quantitatively predict CaCO 3 mineral formation in seawater across a broad range of S A -T values is a major step forward in geochemical modeling.
A limitation, at present, is that the equation for estimating (calcite) is based on seawater S A that works well for seawater compositions, but may not work for other salt assemblages.For example, calcite precipitation in seawater requires high values (high supersaturation) (Figs.1-2), which is due to seawater inhibitory components such as Mg 2+ , orthophosphate, and organic matter (Pytkowicz, 1973;Bischoff et al., 1993;Morse et al., 2007).Other saline waters with different compositions (e.g., exceedingly low or high Mg 2+ , orthophosphate, and organic matter) would likely respond quite differently with respect to calcite precipitation.Another limitation is how to deal with pressure effects.The S A -T data used to estimate the omega values (Fig. 1) were based on a single set of experiments that were done at P =0.1 MPa (Morse and He, 1993).Nevertheless, as formulated in this paper, we can accurately simulate surface properties of seawater, which by itself is a major improvement in understanding seawater chemistry.
Edited by: W. Jenkins

Figure 2 .Fig. 2 .
Figure 2. The pseudo-homogeneous nucleation boundaries for S A -T-P models in Year 2008 (solid lines) and potentially in Year 2100 (dashed lines) Absolute Salinity (g/kg)

Figure 3 .
Figure 3. Model calculated seawater pH values as a function of CO 2 µatm for S A values of 5-45 g/kg at 25°C.

Fig. 3 .
Fig. 3. Model calculated seawater pH values as a function of CO 2 µatm for S A values of 5-45 g kg −1 at 25 • C.

Table 1 .
Initial ion composition of the seawater sample (S A =35.147 g kg −1 ).