Aqueous Solubility Diagrams for Cementitious Waste Stabilization

Waste Management Laboratory, Paul Scherrer Institute,. CH-5232 Villigen PSI, Switzerland, and Geoscience Institute,. Johannes Gutenberg-University, D-...
0 downloads 0 Views 143KB Size
Environ. Sci. Technol. 2002, 36, 2926-2931

Aqueous Solubility Diagrams for Cementitious Waste Stabilization Systems. 4. A Carbonation Model for Zn-Doped Calcium Silicate Hydrate by Gibbs Energy Minimization D M I T R I I A . K U L I K †,‡ A N D M I C H A E L K E R S T E N * ,§ Waste Management Laboratory, Paul Scherrer Institute, CH-5232 Villigen PSI, Switzerland, and Geoscience Institute, Johannes Gutenberg-University, D-55099 Mainz, Germany

A thermodynamic Gibbs energy minimization (GEM) solid solution - aqueous solution (SSAS) equilibrium model was used to determine the solubility of Zn from calcium silicate hydrate (CSH) phases doped with 0, 0.1, 1, 5, and 10% Zn at a unity (Ca+Zn)/Si molar ratio. Both the stoichiometry and standard molar Gibbs energy (Go298) of the Znbearing end-member in the ideal ternary Zn-bearing calcium silicate hydrate (CZSH) solid solution were determined by a “dual-thermodynamic” (GEM-DT) estimation technique. The SSAS model reproduces a complex sequence of reactions suggested to occur in a long-term weathering scenario of cementitious waste forms at subsurface repository conditions. The GEM model of CZSH leaching at several Zn loadings and solid/water (s/w) ratios in a CO2-free system showed that, upon complete dissolution of portlandite and calcium zincate phases at decreasing s/w < 0.01 mol‚kg(H2O)-1, the total dissolved concentrations Siaq, Caaq, and Znaq are controlled by a CZSH solid solution of changing composition, with a trough-like Znaq drop by 2-3 orders of magnitude. Carbonation was simulated in another GEM model run series by CO2 titration of the system with initial s/w ≈ 0.9 mol/kg(H2O). Formation of (Ca,Zn)CO3 nonideal solid solution was predicted already at early reaction stage in the presence of both portlandite and calcium zincate hydrate phases. Upon their disappearance, pH, Znaq, Caq, and fCO2 were predicted to change due to the incongruent dissolution of two concurrent CZSH-I and CZSH-II solid solutions, until the total re-partitioning of Ca and Zn into a carbonate solid solution coexisting with amorphous silica at fCO2 > 0.1 bar. Along this solidphase transition, dissolved Znaq concentrations follow a highly nonlinear trend. The model results predict that at low to moderate Zn loading (e 1% per mole Si), CZSH-type compounds can efficiently immobilize Zn in the near field of a cement-stabilized waste repository.

* Corresponding author telephone: +49-6131-3924366; fax: +496131-3923070; e-mail: [email protected]. † Paul Scherrer Institute. ‡ On leave of absence from State Scientific Centre of Environmental Radiogeochemistry, Palladin Prosp. 34a, 03680 Kyiv, Ukraine. § Johannes Gutenberg-University. 2926

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 13, 2002

Introduction Cements are used in many countries for stabilization and solidification of hazardous wastes containing toxic metals and radionuclides (1). Potential migration of contaminants is ceased by both physical entrapment and chemical binding in cement phases (2). However, cementitious materials are prone to weathering, whereby leaching, secondary phase precipitation/dissolution, and carbonation were determined as a generic reaction sequence (3, 4). Carbonation can markedly alter the microstructure of a hydrated cement due to precipitation of carbonates in the pore spaces, thus reducing leaching of waste elements by drastically slowing down mass transport out of the waste form (5-7). Chemical changes induced by carbonation, however, may also affect the element binding mechanisms into the cementitious waste form, albeit these mechanisms are less well-understood. While calcite seems to appear always as an important reaction end product, mechanisms of intermediate phase formation, and, in particular, chemical fate of entrained metals remain ambiguous due to phase characterization problems. In this paper, the effects of leaching and carbonation will be modeled using the Gibbs energy minimization (GEM) thermodynamic equilibrium approach to simulate the reaction sequence in weathering of (Zn-doped) calcium silicate hydrate (abbreviated CSH as common in cement chemistry) as the predominant phase in Portland cements. In the present contribution, we will show how the solid solution-aqueous solution (SSAS) system for the pure CSH binary solid solution phase described earlier (8) can now be extended to a ternary one by introducing another endmember for a minor metal. So far, solubility data has been provided only for Zn, which point to a solid solution formation of a Zn-bearing CSH phase (abbreviated CZSH; 9-13). A first approximation in modeling such a ternary SSAS thermodynamic equilibrium is seeking for the third end-member stoichiometry and G°298 values consistent with a Raoultian ideal mixing behavior. The ideal SSAS approach appears preferable for an extension to a ternary SSAS system in case of the lack of enough experimental data to derive at empirical ternary interaction parameters needed for any nonideal SSAS model approach. The challenge is that for the Zn-bearing end-member, both stoichiometry and G°298 values must be determined from the available solubility and structural data. The aims of this paper are therefore as follows: (i) to develop a thermodynamic model of a reliable ternary SSAS system, which predicts weathering reactions typical for cementstabilized waste in contact with groundwater; (ii) to test how the new model predicts equilibrium pH and total dissolved Znaq, Caaq, and Siaq concentrations in a range of solid Zn/Ca, Ca/Si, and s/w ratios; and (iii) to ultimately quantify possible Zn release from cementitious waste in contact with CO2-rich groundwater (weathering scenario).

Materials and Methods GEM thermodynamic modeling has been performed using the GEM-Selektor code (http://les.web.psi.ch/Software/ GEMS-PSI), a direct solver of complex SSAS equilibria. The underlying GEM convex-programming approach is described elsewhere (8, 14). Briefly, GEM finds chemical potentials of elements (dual solution, ui) and mole quantities of species (prime solution, xj) that minimize the total Gibbs energy function of the system G(x) at temperature T and pressure P of interest, subject to mass balances of chemical elements and charge. Input data include the system bulk elemental composition b ) {bi} in total moles of elements and zero 10.1021/es010250v CCC: $22.00

 2002 American Chemical Society Published on Web 05/30/2002

TABLE 1. Bulk Compositions of the Aquatic “Basic Subsystems” Used in GEM-DT Calculationsa sample

CaO (mol 10-3)

SiO2 (mol 10-3)

ZnO (mol 10-6)

pH calcd

CSH “pure” CSH 0.1% Zn CSH 1% Zn CSH 5% Zn CSH 10% Zn

1.8 1.8 0.7 0.6 0.5

0.13 0.13 0.3 0.2 0.1

0 0.01 0.06 2.0 10

11.49 11.49 11.00 10.97 10.92

a

1 kg of H2O, 4 mol of N2, and 1 mol of O2 were added in all cases.

charge and the standard molar Gibbs energy functions gj,T,P of aqueous species, gases, solids, and solid solution endmembers. Chemical elements and charge are indexed with i ∈ N, and species (grouped into phases) are indexed with j ∈ L. Thermodynamic data in the system Ca-Zn-N-CSi-O-H-charge are given in Table S-1 of the Supporting Information. The aqueous ion association model is based on the SUPCRT data (15). Nitrogen gas and inorganic carbon species were included to simulate carbonation of the SSAS system under the N2-O2-CO2 atmosphere. Individual activity coefficients of aqueous species were calculated using the extended Debye-Hu ¨ ckel equation in Truesdell-Jones form with Kjelland ion-size parameters and common third parameter. Note that in the aquatic CZSH system, ionic strength is relatively low (I < 0.03 m), hence the impact of the third parameter was minor. Thermodynamic data for solids were compiled from literature (16, 17) and are listed in Table S-2 of the Supporting Information. Solubility of Zn-Doped CSH Samples. Solubility experiments carried out to determine the binding mechanisms of Zn in a CSH matrix were reported previously (9). Zn was coprecipitated in CSH at fractions of 0.1, 1.0, 5.0, and 10 mol % in exchange with Ca at (Ca+Zn)/Si ) 1.0. Total dissolved Caaq, Siaq, and Znaq were measured as a function of pH. In presence of the coprecipitated phase, Znaq was significantly lower than predicted by equilibrium with either ZnO or calcium zincate solids but directly related to Zn concentration in the solid phase. Hence, solid solution formation was assumed to be a predominant intercalation mechanism for Zn in CSH (9). Previous GEM SSAS modeling of the binary CSH system demonstrated that CSH gels at Ca/Si > 0.9 are prone to incongruent dissolution (8). A simple decrease in s/w ratio causes a significant decrease of Ca/Si ratio in solid. This effect becomes even more pronounced if the system is acidified. A layer (of unknown thickness) of a reprecipitated CSH with unknown Ca/Si ratio, different from that in the initial synthetic samples, could have formed when metastable equilibrium has been reached in the experiments with addition of acid. Therefore, only those five experiments without additions of acid or base were selected (Table S-3 in Supporting Information) and converted into bulk compositions as model inputs of five SSAS subsystems (Table 1). End-Member Components in the CaO-ZnO-SiO2-H2O System. Good fits by an ideal SSAS model to a particular set of Ca and Si solubility data can be obtained from reliable assumptions about structure and stoichiometry of CSH endmembers (8). Two ideal binary solid solution phases are to be considered to cover the complete compositional range: CSH-I covers the region of Ca/Si < 0.8 and is important only for extended weathering scenarios, and CSH-II existing at 0.8 < Ca/Si < 1.7 represents the range of CSH compositions to be found in hydrated Portland cements. The structural ideas permit us to write formulas of the (on average) pentameric CSH-II end-members as 4Ca(OH)2‚5SiO2‚4H2O (denoted as T5) for the low Ca/Si ratio end-member and 10Ca(OH)2‚5SiO2‚4H2O (denoted as J5) for the high Ca/Si

end-member. In more mature, fully polymerized CSH-II phases, the hydrated precursors of 14-Å tobermorite and jennite minerals, 5Ca(OH)2‚6SiO2‚5H2O (Tob-II) and 10Ca(OH)2‚6SiO2‚6H2O (Jen), respectively, appear to be more suitable as end-members (8). GEM modeling in this system works well with numerically simplest Raoultian ideal CSH solid solutions, if the number of basic silica and portlandite formula units is expressed using coefficients nSi and nCa (for instance, nSi ) 6 and nCa ) 10 for the Jen end-member). Good model fits to solubility data have been found just by varying both nSi and nCa factors proportionally, i.e., by adjusting the ideal end-member stoichiometries (8). Phase characterization of X-ray amorphous, fully hydrated Zn-bearing CZSH precipitates is difficult, but mechanisms of Zn incorporation by solid solution into CSH have been elucidated previously using X-ray absorption spectroscopy (13). In the defective, domain-like layered structure, distorted ZnO2(OH)24- tetrahedra are likely to replace “bridging” SiO44tetrahedra within (pentameric) “dreierketten” silicate chains. This leads to the basic stoichiometries of {Ca4H4Si4O14Zn(OH)2}‚(H2O)m and {Ca4H4Si4O14Zn(OH)2}(OH)8Ca6‚(H2O)m of Zn-substituted low-Ca (tobermorite) and high-Ca (jennite) compounds, respectively (13). These synthesized CZSH components, however, must not necessarily represent pure Zn-bearing end-members in the ternary SSAS systems. A hint of what such an end-member stoichiometry may consist of was found by heating the samples to 1000 °C. Both hardystonite (Ca2ZnSi2O7) and clinohedrite (CaZnSiO4‚H2O) were found by X-ray diffractometry and infrared spectrometry to form upon CZSH dewatering (18). Both phases, and hemimorphite (Zn4Si2O7(OH)2‚H2O) or junitoite (CaZn2Si2O7‚H2O) as well, are natural silicates structurally similar to tobermorite. It is not yet clear to what extent these minerals can mix within solid solution series and if their fully hydrated analogues exist at room temperature. In the latter case, more Zn can probably enter the CSH structure than that prescribed solely by the Zn-to-Si substitution mechanism, up to a total replacement of Ca (hemimorphite). Thus, we considered for a first approximation all four minerals as potential candidates for a pure Zn-bearing end-member. At Zn loadings below 10 mol %, Zn end-members might therefore be represented by (i) Zn-Si-substituted pentameric calcium zincate silicate hydrate of stoichiometry SiO2‚Ca(OH)2‚[Zn(OH)2]0.25‚H2O or CZ0.25SH; (ii) hardystonite (Hard, SiO2‚Ca(OH)2‚[Zn(OH)2]0.5 or CZ0.5SH); (iii) clinohedrite (Clnh, SiO2‚Ca(OH)2‚Zn(OH)2 or CZSH); (iv) junitoite (Junit, SiO2‚[Ca(OH)2]0.5‚Zn(OH)2 or C0.5ZSH); (v) hemimorphite (Hemim, SiO2‚[Zn(OH)2]2 or Z2SH); or (vi) even a hypothetical pure zinc silicate (ZnSiO3 or ZSH). Chemography of the ZnO-CaO-SiO2-H2O system (Figure S-1, Supporting Information) shows that other endmember compounds are, in principle, also possible, albeit of less probability. Relevant solid phases in that system also include amorphous silica (SH), portlandite (CH), zinc hydroxides (ZH), zincite (Z), and calcium zincate hydrate (CHZ). The latter has been identified as a metastable Zn-bearing phase common in cement systems (13). Unfortunately, thermodynamic data for the potential Zn-bearing CZSH endmembers are lacking. A thermodynamic data evaluation technique called “dual-thermodynamic calculation” was therefore applied to estimate these data from the available stoichiometries and solubilities of these compounds. Moreover, this technique helps to select the most probable candidate for the Zn-bearing CZSH system end-member. Dual-Thermodynamic Calculations. Prime (xj, j ∈ L) and dual (ui, i ∈ N) solutions of the GEM equilibrium problem can be used in dual-thermodynamic (DT) calculations to retrieve the unknown molar free Gibbs energies (gT,P) of solid solution end-members from known bulk compositions of coexisting aqueous and solid phases (14, 19). The DT approach consists of splitting up the equilibrium system of VOL. 36, NO. 13, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2927

TABLE 2. Estimates of Molar Chemical Potentials of Chemical Elements u′i (kJ mol-1) at 1 Bar and 25 °Ca z

sample CSH “pure” CSH 0.1% Zn CSH 1% Zn CSH 5% Zn CSH 10% Zn

uCa

uSi

uZn

-673.356 -863.560

0

uH

uO

-117.593 -1.995

0.001 -673.356 -863.560 -335.702 -117.593 -1.995 0.01 -680.895 -858.026 -328.463 -117.593 -1.995 0.05 -681.658 -858.784 -319.557 -117.593 -1.995 0.1

-682.578 -860.213 -315.313 -117.593 -1.995

a Calculated using the GEM model for the bulk compositions given in Table 1. Example: at z ) 0.1, calculation of the chemical potential of hydrated clinohedrite-like SiO2‚Ca(OH)2‚Zn(OH)2 end-member using eq 1 is given by µ ) uSi + uCa + uZn + 6uO + 4uH ) -(682.578 + 860.213 + 315.313 + 11.97 + 470.372) kJ mol-1 ) -2340.446 kJ mol-1.

interest into a basic and a nonbasic subsystem. The basic subsystem of known bulk chemical composition (vector b′) includes aqueous and other phases with known molar Gibbs energy (gTP) values for all species. Phases and species with unknown gTP values (or activity coefficients) but with known concentrations of species are put into the second nonbasic subsystem. An evidence is essential that both subsystems coexist in equilibrium at T and P of interest (19). This is to be performed along the basic rationale of equilibrium thermodynamics, which is that the chemical potential of a component (e.g., chemical element or electrical charge) must be the same in all coexisting equilibrium phases. The DT technique exploits this classic idea. First, GEM calculation of equilibrium in the basic subsystem only is performed. It results in a dual vector of elemental chemical potentials (u′i) used for computing chemical potentials of species in the nonbasic subsystem: /

µj )

∑a u′ ji

(1)

i

i

Equation 1 alternatively calculates chemical potentials usually defined as µj ) µj° + ln Cj + ln λj through the standard-state chemical potential µj° ) gTP/(RT), concentration Cj, and activity coefficient λj. Substitution of eq 1 for µj and rearrangement yields a DT equation for an end-member of the multicomponent solid solution (14):

gj,TP* ) RT(

∑a u′ - ln λ - ln X ) ij

i

j

j

(2)

i

where g*j,TP estimates the standard-state molar Gibbs energy function of the jth end-member at T,P of interest; Xj is a (known) mole fraction; λj is a known activity coefficient (λj ) 1 in Raoultian ideal solid solutions); R ) 8.314 J K-1 mol-1; and aji stands for the ith element stoichiometry coefficient in chemical formula of the jth species. If n experimental solubility points are available at various solid compositions, the DT-estimated g*j,TP values can be used for selection of an optimal stoichiometry and gT,P value of the sought endmember. A simple selection criterion is the smallest standard deviation (arg min s(g*)), where s is an uncertainty estimate for gj,TP. Supporting DT calculations for major solid solution end-members may help in assessing reliable standard deviations of g°j,T for the sought Zn-bearing end-member. The Ca-Zn-Si-O-H-N-charge thermodynamic system for GEM-DT calculations has been composed of the H2O solvent, aqueous species, SH, CH, and CHZ phases (see Supporting Information, Tables S-1 and S-2). GEM runs for five basic subsystems (Table 1) at P ) 1 bar and T ) 25 °C yielded the dual chemical potentials u′i listed in Table 2. 2928

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 13, 2002

However, application of eq 2 also requires mole fractions Xj. In the case of CZSH, Xj values depend on the choice of Zn end-member stoichiometry. Denoting the bulk mole fraction of Zn(OH)2 with z, composition of any CZSH solid can be written as SiO2‚[Ca(OH)2]1-z‚[Zn(OH)2]z at unity molar ratio of (Ca+Zn)/Si. Hence, normalization to one SiO2 formula unit greatly simplifies determination of Xj from the bulk solid compositions. Taking tobermorite with a moles of Ca(OH)2 per SiO2, and jennite with b moles of Ca(OH)2 per SiO2 as binary CSH end-members, the stoichiometry of a Zn endmember with d moles Zn(OH)2 becomes SiO2‚[Ca(OH)2]c‚ [Zn(OH)2]d. Denoting mole fractions of three constituent (hydr)oxides as X1, X2, and X3, respectively, one can write:

SiO2‚(1 - z)Ca(OH)2‚zZn(OH)2 ) X1{SiO2‚aCa(OH)2} + X2{SiO2‚bCa(OH)2} + X3{SiO2‚cCa(OH)2‚dZn(OH)2} (3) The SiO2 unit can be dropped from eq 3 because X1 + X2 + X3 ) 1, which leads to the constraints aX1 + bX2 + cX3 ) 1 - z, dX3 ) z, X2 ) 1 - X1 - X3, and, after some algebraic rearrangements, to X1 ) [1 - b + z/d(b - c - d)]/(a - b), X2 ) [1 - a + z/d(a - c - d)]/(b - a), and X3 ) z/d. The mole fraction X3 of the Zn-bearing end-member does not depend on the stoichiometry of tobermorite and jennite endmembers. The above constraints can now be used for deriving Xj values for eq 2 from known Zn mole fraction z in bulk Zn-doped samples. These Xj values are given in Table 3, together with the thus DT-calculated molar Gibbs energies g*j,298 ) RTµ*j of all potential Zn end-member candidates.

Results DT calculation results (Table 3) for the six Zn-containing end-member candidates show a pronounced minimum of standard deviation s(g*) (Figure 1). Only twosCZSH and CZ0.5SHshave s(g*) < 3 kJ mol-1 ( 1. The pentameric endmembers (C0.8SH and C1.6SH) did not yield smaller s(g*) values and have not been further considered. Solid solution end-members are not confined to pure naturally occurring analogues. Selection of other rational stoichiometries may result in even better fits to solubility data than obtained in DT calculations for the clinohedritelike end-member. Note, however, that the thermodynamic “DT-arg min s“ procedure is not confined to structural arguments but could use a grid of possible end-member stoichiometries, compute mean values of g* and s over n solubility points, and plot isolines of s(g*) on a chemography diagram. In this way, optimal end-member stoichiometries together with g*T,P values and uncertainties can be derived algorithmically. For instance, using data from Table 2 for C0.5Z0.5SH and C0.75Z0.75SH stoichiometries yields a mean g*298 ) -1596.75 kJ mol-1 with s(g*) ) 1.694 kJ mol-1 and g*298 ) -1966.40 kJ mol-1 with s(g*) ) 1.253 kJ mol-1, respectively,

TABLE 3. Dual-Thermodynamic (DT) Calculations of g*298 and s(g*298) (in kJ mol-1) for Six Possible Stoichiometries of Fully Hydrated Zn-Bearing CZSH End-Member end-member

X3 (z ) 0.001)

g*298 (z ) 0.001)

X3 (z ) 0.01)

g*298 (z ) 0.01)

Zn-tobermorite, CZ0.25SH hardystonite, CZ0.5SH

0.004 0.002

-2147.296 -2052.116

0.04 0.02

-2153.199 -2056.209

clinohedrite, CZSH junitoite, C0.5ZSH hydrated ZnSiO3, ZSH hemimorphite, Z2SH

0.001 0.001 0.001 0.0005

-2337.837 -1881.570 -1425.304 -1998.464

0.01 0.01 0.01 0.005

-2338.311 -1878.275 -1418.239 -1984.160

b

X3 (z ) 0.05)

g*298 (z ) 0.05)

X3 (z ) 0.1)

g*298 (z ) 0.1)

0.2 0.1

-2156.483 -2057.266

0.4 0.2

-2159.489 -2059.211

0.05 0.05 0.05 0.025

-2334.914 -1874.497 -1414.080 -1971.095

0.1 0.1 0.1 0.05

-2334.738a -1873.860 -1412.983 -1965.754

g*298 average

s(g*)

-2154.117 5.222 -2056.201 2.994 -2057.945b 1.113b -2336.450 1.886 -1877.051 3.588 -1417.652 5.581 -1979.868 14.611

a Example using eqs 2 and 4 and Table 2: g*(CZSH) ) µ(CZSH) - RT ln X -1 ) -2334.738 kJ mol-1. CZSH ) (-2340.446 - 2.479) × -2.3026 kJ mol Calculated from three points only, omitting a point at z ) 0.001.

FIGURE 1. Results of DT selection (see Table 3 and text for explanations). i.e., only 0.2-0.3 pK unit uncertainty. Such refinements, however, would require many more (n > 7) solubility data points distributed over the whole compositional space not yet available for evaluation of the quality of various ideal and even nonideal solid solution models.

Discussion Modeling the Impact of Leaching on CZSH. Leaching of Zn from cement matrix by carbonate-free water can now be simulated in model runs of sequences of SSAS equilibria at varying Zn in CSH loadings, controlled by increasing solid/ water (s/w) mass ratio in bulk composition of the Ca-ZnSi-O-H system. Leaching results in a preferential release of Ca into aqueous phase from the CSH solid solution phases (8). In batch simulations, the initial Ca/Si ratio would decrease and pass unity. At this point the quality of DT g*T,P values for the ideal CZSH end-member can be assessed from available experimental solubility data (9). Initial bulk composition of the system included 1 kg of H2O, 4 mol of N2, 1 mol of O2, and 1 mol of solid SiO2‚(2 z)Ca(OH)2‚zZn(OH)2. Four process simulations at z ) 0.001, 0.01, 0.05, and 0.1 were run with stepwise change of log(s/w) from 0 to -4.2, with step -0.2 until complete disappearance of solid phases. The aqueous, gaseous, and relevant solid phases (CHZ, CH, ZH, and SH) were included, plus two ternary ideal CZSH solid solutions. The phase with Tob-II, Jen, and Hds-II end-members with DT-estimated g*298 values is now called CZSH-II; another phase with SH, Tob-I, and Hds-I end-members is called CZSH-I analogous to the CSH-I phase (8). The Hds-II and Hds-I end-members were assumed to have the same stoichiometry and g°298 values (Table 4). A hypothetical CZSH-I phase was introduced to cover the silicarich region at Ca/Si ratios less than that of tobermorite ( 0.01 mol (kg of H2O)-1, pH, Caaq, and Znaq are fixed by the presence of excess CH and CHZ. The latter two phases disappear at s/w < 0.01, where Caaq and Znaq become controlled only by the CZSH-II phase of drastically changing composition (Figure 2b). Note that this change in solid solution composition coincides with a trough-like Znaq minimum of 2-3 orders of magnitude (Figure 2a). At different total Zn loadings, the Znaq minima are located at s/w ≈ 0.001 at the maxima in mole fraction of Tob-II end-member (Figures 2b and 3). Experimental SSAS data (Table 1 and error bars in Figures 2 and 3) are close to these minima, which indicates consistent behavior of the CZSH ideal solid solution model in all GEM model runs. Further decrease in the s/w ratio results in complete dissolution of CZSH-II phase at s/w < 0.0001. Calculations with the alternative clinohedrite Zn end-member SSAS produced similar plots but with somewhat less pronounced minima in Znaq (Figure 3b). Evidently, the choice of Zn-containing end-member stoichiometry leads to different trends of Znaq at trace Zn loadings. For instance, at z ) 0.001, the hardystonite-like end-member model (Figure 3a) predicts about 2 orders of magnitude lower Znaq than that of clinohedrite (Figure 3b). At present, experimental data are insufficient to decide which of both end-members is more reliable. Nonetheless, these results clearly demonstrate the importance of the s/w ratio (i.e., the ratio of actually reacted solid to aqueous phase mass) in modeling metal leaching in cementitious repository systems, which do not follow a simple linear dissolution/precipitation relationship as often assumed. Modeling the Impact of Carbonation on CZSH. Under subsurface waste repository conditions, hydrated cements would react with CO2-containing groundwater resulting in an incongruent dissolution of CH and CSH and precipitation of calcium carbonate. In this typical concrete weathering scenario (7, 20, 21), any intercalated metal can be partially mobilized into aqueous solution, precipitate into its own oxide or carbonate (or sulfide in anoxic waters) phase, or get incorporated into a solid solution with CaCO3. GEM calculations can easily reproduce such chemical mass transfer, helping in (at least qualitative) prediction of the possible metal re-partitioning between the near-field groundwater, CSH and carbonate solid solutions, or other solid phases. A regular nonideal (Ca,Zn)CO3 solid solution between calcite and smithsonite end-members has been added to the model system definition. A Guggenheim mixing parameter a0 ) 2.3 was estimated using Lippmann’s approach and corrections thereof described earlier (14), together with carbonate aqueous species and gases compiled in Tables S-1 and S-2 of the Supporting Information. The bulk system composition was set by adding 0.9 mol of CZSH stoichiometry, SiO2‚(2-z)Ca(OH)2‚zZn(OH)2 at z ) 0.001, 0.01, 0.05, and 0.1. Evolution of this SSAS system at constant s/w has been simulated by titration with CO2 at a stepwise log(CO2) addition from -0.1 to -2.4 at increment -0.01 (in log mol). VOL. 36, NO. 13, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2929

TABLE 4. Stoichiometry, Standard Molar G°298 Values, and Solubility Products of Potential CZSH End-Membersa end-member

stoichiometry

G°298 (kJ mol-1)

log K298

amorph. silica, SH-I tobermorite, Tob-I tobermorite, Tob-II jennite, Jen hardystonite, Hds-I,II clinohedrite, Cln-I,II

x ) 1, y ) 0, w ) 0 x ) 2.4, y ) 2, w ) 2 x ) 1, y ) 0.83, w ) 0.83 x ) 1, y ) 1.667, w ) 1 x ) 1, y ) 1, z ) 0.5, w ) 0 x ) 1, y ) 1, z ) 1, w ) 0

-849.45 ( 1.7b -4383.77 ( 11c -1826.57 ( 3.3c -2626.41 ( 10c -2057.57 ( 3.0c,d -2336.45 ( 1.9c

-1.20e 25.42f 10.59f 27.35f 19.94f 25.54f

a Solid stoichiometry is expressed as xSiO ‚yCa(OH) ‚zZn(OH) ‚wH O. b From ref 15. c This work; for Tob and Jen, the uncertainty estimates 2 2 2 2 are taken as differences between DTG values and (Table 4 in ref 8). d Weighted average of two g*298 estimates from Table 3. e Reaction: SiO2(quartz) f + 2+ 2+ 0 ) SiO2(SH). Reaction: solid + (2y + 2z)H ) xSiO2 + yCa + zZn + (2y + 2z + w)H2O, data for aqueous species from Table S-1 (Supporting Information).

FIGURE 2. Example of GEM model simulation of leaching at 1% Zn in CSH loading, z ) 0.01 (CZSH-II solid solution with hardystonite end-member). (a) Total aqueous solubilities Caaq, Znaq, Siaq, and pH where bars at a s/w ratio of about 10-3 correspond to the uncertainty intervals of experimental solubility data (9), positioned along the abscissa according to the solid-phase composition. (b) Mole fractions of CZSH-II end-members and presence of single-component solids; hexagonal dots correspond to the solid-phase composition in the solubility experiment for 1% Zn in CSH loading. A Ca-Zn carbonate phase with small XZn was predicted to occur throughout the whole SSAS equilibrium profile. In presence of excess portlandite, dissolved carbonate with Caq ≈ 10-5 m at fCO2 ≈ 10-13 bar was computed. Figure 4 shows GEM model run results for z ) 0.01 with hardystonite endmembers in both CZSH phases. On further CO2 addition to the system, excess CH and CHZ phases disappear, and the values of pH, Znaq, Caq, and fCO2 drastically change due to a complex incongruent dissolution of concurrent CZSH-I and CZSH-II solid solution phases. This development ultimately ends up in a total re-partitioning of Ca and Zn into the carbonate solid solution coexisting with almost pure amorphous silica at a fCO2 > 0.1 bar. Concurring with this quite complex phase precipitation/dissolution pattern, Znaq 2930

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 13, 2002

FIGURE 3. GEM model simulation runs of leaching at 1 bar, 25 °C, dependence of Znaq on Zn loading (z in %, numbers at curves): (a) with hardystonite-like, (b) with clinohedrite-like CZSH-II endmember. The bars at a s/w ratio of about 10-3 correspond to the uncertainty intervals of experimental solubility data (9). curves with minima at 10-8 to 10-6.5 m were predicted, quite similar in shape to those in previous (CO2-free) leaching models (Figure 2). These mimima, in turn, coincide with the maxima in mole fractions of Tob-I or Tob-II end-members at about 0.1 mol of CO2/kg of H2O. In another model run, but at higher Zn loading (z ) 0.1) and with clinohedrite-like end-members in both CZSH phases (Figure S-2, Supporting Information), portlandite did not appear at all because all excess calcium was kept in CHZ phase already at very early stages of CO2 titration. The model produced a Znaq concentration curve again with a minimum at about 10-6 m. This time, however, excess zincite (ZnO) appears together with CZSH-I at intermediate process stages (log CO2 ≈ -0.75), just before the ultimate change to Ca-Zn carbonate-dominated (XZnCO3 < 6%) SSAS equilibrium. Model Implications. A challenging question was to what extent CZSH phases can control Zn retention in cementitious waste repositories. Our GEM model predicts a weathering sequence, in which the preferential removal of Ca from CH, CHZ, CSH, and CZSH phases is a key factor of reaction progress, both in the presence and absence of CO2. Progressive incongruent dissolution during carbonation of CH, CHZ,

in more experimental and theoretical detail. Effects of coprecipitation of other major elements (Al, Mg, Fe, Na, K, sulfate) were beyond this paper but can potentially influence immobilization of Zn by introducing competitive sorption phases or changing calcium solubility. Extension of this thermodynamic GEM model to other elements of environmental concern (especially radionuclides) appears quite promising for long-term weathering scenario simulations.

Acknowledgments This work was funded in part by the European Science Foundation (ESF GPoll Exchange Grant CRP99/02), the German Science Foundation (DFG Grant Ke508/5), and the Swiss National Co-operative for Nuclear Waste Disposal (NAGRA). The authors are grateful to Urs Berner and Caterina Tommaseo for stimulating discussions. Thoughtful comments by Pierre Glynn were quite helpful in improving the final manuscript.

Supporting Information Available Standard (partial molal) thermodynamic properties of aqueous species and solids used in the GEM calculations, together with a compilation of experimental data used for model verification. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited

FIGURE 4. Example GEM simulation of a weathering scenario (CO2 titration at 1 bar, 25 °C) involving a (Ca,Zn)CO3 nonideal binary, CZSH-I and CZSH-II ideal ternary solid solutions with a hardystonitelike end-member. (a) Total aqueous concentrations, pH, and CO2 gas fugacity. (b) Composition of solid phases. and CZSH phases leads ultimately to formation of Ca/Zn carbonate solid solution even at constant total s/w ratio. The weathering process can be followed in a reaction sequence where (i) in the presence of excess calcium hydroxide (Ca/Si > 1.2-1.5), Znaq ≈ 10-5 m is set by CHZ solubility; (ii) at moderate Ca/Si ratios (0.9-1.2), Znaq is determined by stability of CZSH-II ternary solid solution as function of its composition, Zn loading, and s/w ratio; and (iii) minimum Znaq concentrations (10-9-10-6 m, depending on Zn loading) to occur at 0.8 < Ca/Si < 1.1 (pH ≈ 11) and at the largest mole fraction of tobermorite end-members in coexisting CZSH-I and CZSH-II phases. If carbonation extends to Ca/Si < 0.8 in CZSH-I phase, Znaq increases to 10-5-10-6 m, controlled by zincite or (Ca,Zn)CO3 solid solution. CSH with 0.8 < Ca/Si < 1.2, and Zn/Ca e 0.1, can be deduced as optimal for the long-term Zn fixation from the results (Figures 3 and 4). All model runs predict an increase in Zn solubility by 2 orders of magnitude in a narrow region of existence of silicapoor CZSH-I phase, where fCO2 changes from < 10-6 to 10-1 bar, i.e., within the typical groundwater range. Thus, design of cement-stabilized waste repositories must also consider a possible sharp increase in Znaq (up to 10-5 m) at too low s/w ratios, or too high fCO2. To approach real repository conditions, however, the kinetics of incongruent CZSH dissolution must be addressed

(1) Spence, R. D., Ed. Chemistry and microstructure of solidified waste forms; Lewis: Boca Raton, FL, 1993. (2) Glasser, F. P. J. Hazard. Mater. 1997, 52, 151. (3) Berner, U. R. Waste Manage. 1992, 12, 201. (4) Reardon, E. J. Waste Manage. 1992, 12, 221. (5) Neall, F. B. Mater. Res. Soc. Symp. Proc. 1996, 412, 483. (6) Pfingsten, W.; Shiotsuki, M. Mater. Res. Soc. Symp. Proc. 1998, 506, 805. (7) Venhuis, A.; Reardon, E. J. Environ. Sci. Technol. 2001, 35, 4120. (8) Kulik, D. A.; Kersten, M. J. Am. Ceram. Soc. 2001, 84, 3017. (9) Johnson, C. A.; Kersten, M. Environ. Sci. Technol. 1999, 33, 2296. (10) Tommaseo, C. E.; Kulik, D. A.; Kersten, M. In Applied Mineralogy; Rammlmair G., Ed; Balkema: Amsterdam, 2000; pp 919-921. (11) Ziegler, F.; Scheidegger, A. M.; Johnson, C. A.; Daehn, R.; Wieland, E. Environ. Sci. Technol. 2001, 35, 1550. (12) Rose, J.; Moulin, I.; Masion, A.; Bertsch, P. M.; Wiesner, M. R.; Bottero, J.-Y.; Mosnier, F.; Haehnel, C. Langmuir 2001, 17, 3658. (13) Tommaseo, C. E.; Kersten, M. Environ. Sci. Technol. 2002, 26, 0000-0000. (14) Kulik, D. A.; Kersten, M.; Heiser, U.; Neumann, T. Aquat. Geochem. 2000, 6, 147. (15) Shock, E. L.; Sassani, D. C.; Willis, M.; Sverjensky, D. A. Geochim. Cosmochim. Acta 1997, 61, 907 (http://zonvark.wustl.edu/ geopig). (16) Robie, R. A.; Hemingway, B. S. U.S. Geol. Surv. Bull. 1995, No. 2131. (17) Babushkin, V. I.; Matveyev, G. M.; Mchedlov-Petrossyan, O. P. Thermodynamics of Silicates; Springer-Verlag: Berlin, 1985. (18) Tommaseo, C. E. EXAFS-Untersuchungen zur Rolle von Silicium bei der Sorption von umweltrelevanten Schwermetallen (Zn, As, Pb) in Speichermineralen (FeOOH, CSH) (EXAFS analysis of the role of silicium in sorption by reservoir minerals (FeOOH, CSH) of heavy metals (Zn, As, Pb). Ph.D. Thesis, Johannes Gutenberg-University, Mainz, 2002 (in German). (19) Karpov, I. K.; Chudnenko, K. V.; Kulik, D. A.; Avchenko, O. V.; Bychinski, V. A. Geochem. Int. 2001, 39, 1108. (20) Lange, L. C.; Hills, C. D.; Poole, A. B. Environ. Sci. Technol. 1996, 30, 25. (21) Walton, J. C.; Bin-Shafique, S.; Smith, R. W.; Gutierrez, N.; Tarquin, A. Environ. Sci. Technol. 1997, 31, 2345.

Received for review October 1, 2001. Revised manuscript received April 17, 2002. Accepted April 29, 2002. ES010250V

VOL. 36, NO. 13, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2931