Environ. Sci. Technol. 2003, 37, 2184-2191
EXAFS Study of Sn(IV) Immobilization by Hardened Cement Paste and Calcium Silicate Hydrates I S A B E L L E B O N H O U R E , * ,† ERICH WIELAND,† A N D R EÄ M . S C H E I D E G G E R , † , ‡ MICHAEL OCHS,§ AND DOMINIK KUNZ† Waste Management Laboratory and Swiss Light Source, Paul Scherrer Institute, 5232 Villigen, Switzerland, and BMG Engineering LTD, 8952 Schlieren-Zu ¨ rich, Switzerland
In this study, the immobilization mechanisms of Sn(IV) onto calcium silicate hydrates (C-S-H) and hardened cement paste (HCP) have been investigated by combining wet chemistry experiments with X-ray absorption spectroscopy (XAS). Evidence is presented which demonstrates the formation of a Sn(IV) inner-sphere surface complex on C-S-H with a CaO/SiO2 weight ratio of 0.7. Two possible structural models, implying a corner sharing between the Sn octahedra and Q1 or Q2b Si tetrahedra, have been developed based on the experimentally determined structural parameters. In HCP, the formation of a different type of Sn(IV) inner-sphere complex has been observed, indicating that C-S-H may not be the uptakecontrolling phase for Sn(IV) in the cement matrix. An alternative structural model for Sn(IV) binding in HCP has been developed, assuming that ettringite is the uptakecontrolling phase. At high Sn(IV) concentrations, Sn(IV) immobilization in HCP occurs due to the formation of CaSn(OH)6.
Introduction Cement-based materials play an important role in multibarrier concepts developed worldwide for the safe disposal of radioactive wastes in underground repositories. Cement is used to condition the waste materials, for engineered barriers, and for construction of the repositories. The nearfield of cementitious repositories acts as a chemical barrier for radionuclides, thus retarding migration of waste ions into the surrounding geosphere (far-field). Hydrated cement is a complex mixture of calcium silicate hydrates (C-S-H phases), portlandite, and calcium aluminates (1). Generally, C-S-H phases, which are among the chemically most stable cement minerals under highly alkaline near-field conditions, seem to be important for the retardation of waste ions (2). Among the radionuclides of interest, the long-lived fission products have to be taken into consideration because of their contribution to radiotoxicity in the long term. In particular, the long half-life of 126Sn (t1/2 ) 105 y) implies that Sn could have a significant effect on the cumulative radioactive dose if transported without retardation through the geosphere, and therefore, its retention behavior has been studied (3). In * Corresponding author fax: 00 41 56 310 44 38; e-mail: isabelle.
[email protected]. † Waste Management Laboratory. ‡ Swiss Light Source. § BMG Engineering LTD. 2184
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 10, 2003
a cementitious environment (i.e., pH ) 12-14, -400 mV< Eh< +200 mV (4, 5)), Sn is stable in its tetravalent form, Sn(IV), and Sn(OH)5- and Sn(OH)62- are the dominant hydrolytic species (6, 7). To date, very few macroscopic (8) and no spectroscopic studies on the uptake of Sn(IV) by HCP and C-S-H phases exist. Recently, it was shown that the precipitation of CaSn(OH)6 can occur under cement pore water conditions if Sn(IV) is added at concentrations higher than 10-6 M at pH > 12.5 (9). Thus, two possible immobilization processes for Sn(IV) in cement systems are envisaged depending on the Sn(IV) inventory: (1) precipitation of CaSn(OH)6 and (2) Sn(IV) uptake. It should be noted that X-ray absorption spectroscopy (XAS) has been found to be a powerful tool for investigating the binding mechanism of various elements in cementitious systems (10-20). The goal of the present XAS study is (i) to corroborate the formation of Ca[Sn(OH)6] in cement systems; (ii) to determine the local coordination environment of sorbed Sn(IV) in hydrated cement in order to identify the Sn(IV) uptake mode (outer/inner-sphere surface complexation, incorporation, or surface precipitation); and (iii) to identify the Sn(IV) uptakecontrolling phase in the complex cement matrix. Prior to the spectroscopic study, a sorption isotherm for Sn(IV) uptake by HCP was determined to select appropriate Sn/HCP samples for the Sn K-edge XAS measurements. Sn/C-S-H samples were investigated by way of comparison and because C-S-H is an important constituent of the cement matrix.
Materials and Methods Wet Chemistry Studies and XAS Sample Preparation. Sample preparation and wet chemistry studies were carried out in a glovebox under a N2 atmosphere (CO2, O2 < 2 ppm) at room temperature (T, 25 ( 3° C). Hardened cement paste (HCP) was prepared from commercial sulfate-resisting Portland cement (type CPA 55 HTS, Lafarge, France). It was manufactured as reported by Sarott et al. (21) and used as crushed material (22). Two C-S-H gels with target CaO/ SiO2 (C/S) weight ratios of 0.7 and 1.2 were synthesized using a standard procedure (23). According to X-ray diffraction (XRD) data, the materials corresponded to a semicrystalline C-S-H(I) phase (1). Thermogravimetry and quantitative XRD were used to determine the Ca(OH)2 content (C-S-H 0.7, < 1%; C-S-H 1.2, 9%) and to estimate the effective C/S ratios (C-S-H 0.7, C/S ) 0.65 ((10%); C-S-H 1.2, C/S ) 1.00 ((10%)) (24). In the present paper, dissolved X elemental concentrations will be noted [X]aq and total (solid + solution) elemental X concentrations will be noted [X]tot. Prior to use for sample preparation, the C-S-H phases were preequilibrated for at least one week with artificial equilibrium solutions whose compositions were determined from previous inhouse experiments (C-S-H 0.7, [Ca]aq ) 4.7 × 10-5 M, [Si]aq ) 2.1 × 10-3 M; C-S-H 1.2, [Ca]aq ) 1.3 × 10-3 M, [Si]aq ) 5 × 10-5 M). For HCP, an artificial cement pore water (ACW) with a chemical composition corresponding to a solution in equilibrium with a freshly prepared HCP was used ([Na]aq ) 0.114 M, [K]aq ) 0.182 M, [Ca]aq ) 1.6 × 10-3 M, [Si]aq ) 5 × 10-5 M, [Al]aq ) 5 × 10-5 M, and [S]aq ) 3 × 10-3 M). Sn stock solutions were prepared by dissolving SnCl4‚ 5H2O in Milli-Q purified water. A sorption isotherm for Sn(IV) on HCP was determined (solid to liquid (S/L) ratio ) 2.5 × 10-2 kg L-1, 5 × 10-5 M < [Sn]tot < 2 × 10-2 M). The Sn suspensions were shaken end-over-end for 30 days, which was the time required to reach equilibrium, and then centrifuged (60 min at 95 000g). Aliquots were withdrawn from the supernatant and analyzed using ICP-AES (inductively coupled plasma-atomic emis10.1021/es020194d CCC: $25.00
2003 American Chemical Society Published on Web 04/12/2003
TABLE 1. Initial Sn Concentration ([Sn]tot) and Sn Speciation Deduced from EXAFS Measurements [Sn]tot mol‚L-1
equilibration time
Sn speciation
HCP-Ov
10-2
HCP-Int
2 × 10-3
HCP-Und
4 × 10-4
1 day 30 days 1 day 30 days 1 day 30 days 20 × 2 days 1 day 30 days 1 day 30 days
CaSn(OH)6 CaSn(OH)6 CaSn(OH)6 + Sn(1)a CaSn(OH)6 + Sn(1)a Sn(1) Sn(1) n.d.c Sn(2) CaSn(OH)6 + Sn(2)b CaSn(OH)6 CaSn(OH)6
sample
HCP-Rep C-S-H 0.7
1.1 × 10-6 2 × 10-3
C-S-H 1.2
2 × 10-3
a Estimated: 20% Sn(1); 80% CaSn(OH) . 6 30% CaSn(OH)6. c n.d.: not determined.
b
Estimated: 70% Sn(2);
sion spectrometry) or ICP-MS (inductively coupled plasmamass spectrometry). XAS samples are listed in Table 1 together with the initial Sn concentrations used for sample preparation. The samples were prepared by mixing 1 g of crushed HCP or 7 g of wet C-S-H 0.7/1.2 with 40 mL of artificial solutions, which were prepared as described above. Sn stock solution was added in order to reach the following concentrations in the suspensions: [Sn]tot ) 4 × 10-4, 2 × 10-3, and 10-2 M for HCP; [Sn]tot ) 2 × 10-3 M for C-S-H. The suspensions were then equilibrated for 1 day and 30 days. In the following text, HCP samples are named according to the equilibrium conditions with respect to the solubility of CaSn(OH)6 ([Sn]aq ∼ 10-6 M (9)): HCP-Und(ersaturated), HCP-Int(ermediate), HCP-Ov(ersaturated). An additional sample, HCP-Rep(laced), was prepared by mixing 1 g of HCP with 40 mL of a Sn solution ([Sn]tot ) 2 × 10-6 M mixed with ACW and filtered), which was in equilibrium with respect to CaSn(OH)6. After 2 days equilibration, the suspension was centrifuged, and the supernatant was replaced by fresh Sn solution. This procedure was repeated 20 times and allowed progressive accumulation of Sn on HCP without any possible CaSn(OH)6 precipitation. The HCP and C-S-H residual wet solids obtained after centrifugation were sealed with Kapton tape in Plexiglas holders and used for the XAS measurements. Aliquots of the supernatants were analyzed for Sn and Ca using ICP-AES and ICP-MS. CaSn(OH)6 was synthesized and characterized as reported in Lothenbach et al. (9). SnO, SnO2 (Merck, reagent grade chemicals), and CaSn(OH)6 XAS samples were prepared by diluting the solids with cellulose in order to achieve an X-ray absorption edge jump of 1. A reference solution (named Sn4+(OH-)) was prepared by dissolving SnCl4‚5H2O in a NaOH solution of pH 13.3 ([Sn]aq ) 10-3 M). Sn K Edge (29.2 keV) X-ray Absorption. These spectra were collected at the Swiss Norwegian Beam Line (SNBL) at the European Synchrotron Radiation Facility (ESRF) in Grenoble, France, using a channel cut crystal monochromator (Si 111), in transmission (ionization chambers: I0, N2/Kr 90:10; I1,N2/Kr 70:30) and fluorescence modes (Lytle detector; 100% Kr; In filter). A metallic Sn foil was used for energy calibration. The measurements were carried out at room temperature, and the CO2 sensitive samples were placed inside an airtight box that had been flushed with N2. Treatment of the Sn EXAFS data was performed using the set of programs “EXAFS pour le Mac” (25) and a wellestablished method for data analysis (26): preedge extrapolation by a linear function, atomic absorption removal using a sixth order polynomial function, and normalization by the Heitler-Heisenberger method. The origin energy E0 was assigned to the maximum of the absorption edge. The pseudo radial distribution functions (PRDFs) were obtained by
FIGURE 1. [Sn] sorbed on HCP expressed as a function of [Sn]aq after equilibration ((a) 9, sorption isotherm) and as a function of [Sn]tot initially present ((b) b). The HCP XAS samples position on both curves is reported (see “Materials and Methods” for details). Fourier transformation of the k3-weighted extended X-ray absorption fine structure (EXAFS) oscillations (k-range 3.24-14.68 Å-1 or 3.24-10.72 Å-1) using a Kaiser apodization window (τ ) 3). Data fitting was performed using FEFF7 phase and amplitude functions (27) and the “Round Midnight” code (25). This approach was checked on the reference compounds. The fits included only single scattering phenomena, implying that linear chains of three or more atoms are unlikely in these systems, and thus, multiple scattering paths are negligible. This assumption was a posteriori checked by using atomic coordinates of the Sn(2) model (cf. Discussion, § Structural Models for Sn(1) and Sn(2)) to generate a FEFF input file. Possible multiple scattering paths up to the 4th order had negligible amplitudes. The reduction factor, S02, and the mean free path, Γ () k/λ(k)), were used as obtained with FEFF7. The shift between the E0 value fixed in the experimental spectra and the E0 value used by FEFF7 was determined to be 3 eV from simulations of the EXAFS spectra of the reference compounds. A unique additional relative energy threshold (-2 eV < ∆E0 < 2 eV) was fitted for all shells of an adjustment. The fitted kχ(k) functions were obtained by inverse Fourier transform (R range 1.18-2.09 Å for first peak analysis; 1.18-3.85 Å for HCP-Und and C-S-H 0.7). For HCP-Und and C-S-H 0.7 samples, the second peak (2.09-3.85 Å) was fitted separately prior to multi-shell fit (first and second peaks). Floating parameters on the first oxygen shell (1.18-2.09 Å) were the DebyeWaller factor, σ, and the Sn-O distance, R. Floating parameters on further shells (Al, Si, Ca, 2.09-3.85 Å) were σ, R, and the number of neighbors, N. The number of first oxygen neighbors was set to 6, which is the common oxygen coordination number in Sn(IV) compounds. For all samples, the statistical error bars on the fitted parameters were calculated using a procedure described earlier (28). It is noteworthy that only the statistical component (and not the systematic component) of the error was taken into account. With this, errors may be underestimated (29). The structural models proposed in this study were drawn with the CrystalMaker3.0 program.
Results Wet Chemistry. Figure 1a shows the Sn sorption isotherm for HCP, and b shows the initial Sn concentrations added to the suspensions for the determination of the sorption isotherm. The isotherm data show two distinct regions: first, the linear range (slope ) 1) between 1 × 10-8 M < [Sn]aq < 8 × 10-7 M; and second, a sharp increase in Sn bound to HCP due to the formation of a solubility-limiting phase, presumably CaSn(OH)6, above this concentration range (solubility limit [Sn]aq ∼ 1.1 × 10-6 M at pH 13.3) (9). From the above it appears that the experimental window for the preparation VOL. 37, NO. 10, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2185
FIGURE 2. Sn K-edge EXAFS oscillations for Sn reference compounds (SnO2, Sn4+(OH-), and CaSn(OH)6), and for Sn immobilized on hardened cement paste (HCP-Ov/-Int/-Und/-Rep) and on calcium silicate hydrates (C-S-H 0.7/1.2) (___ raw spectra; - - - FEFF; ___ _ 80% CaSn(OH)6 + 20% HCP-Und). Typical features of CaSn(OH)6 are noted A and B. The smaller k-range used for HCP-Und/-Rep and C-S-H 0.7/1.2 is due to a nonremovable glitch at 11 Å-1 (low concentrated samples). The spectrum of cement-rRep could not be exploited due to a nonremovable glitch at 6.2 Å-1 (noted *). of sorption-type XAS samples was extremely narrow, i.e., ranging between Sn sorbed ) 0.01 mol kg-1 (lowest concentration yielding XAS analyzable data at the SwissNorwegian beamline) and Sn sorbed ) 0.03 mol kg-1 (upper bound value limited by the sorption isotherm). At higher Sn loadings, however, contamination of the samples with CaSn(OH)6 was expected. For the preparation of the XAS samples, initial Sn concentrations were selected in such a manner to cover both regions of the isotherm (Figure 1). For the HCP-Ov and HCP-Int samples, the equilibrium Sn concentrations corresponded to the solubility of CaSn(OH)6. For the HCP-Und and HCP-Rep samples, the equilibrium Sn concentrations were found to be below the solubility limit of CaSn(OH)6. X-ray Absorption Spectroscopy. XAS measurements were conducted on Sn(IV)-treated HCP and C-S-H phases (C/S ratios 0.7 and 1.2) (Table 1). The C-S-H phase with C/S ) 0.7 corresponds to a pure “tobermorite-like” C-S-H(I) phase. A relatively pure C-S-H phase with C/S ) 1.2 was synthesized at pH 13.3 and also used for the XAS studies. C-S-H phases with higher C/S ratios, which were synthesized at pH 13.3, were found to be strongly contaminated with portlandite, and therefore, were not used for the present study. It is to be noted that C-S-H phases with C/S ∼ 1.7 are expected to predominate in cement matrixes (1). X-ray Absorption Near Edge Structure (XANES). The Sn K-edge position was examined and a 4 eV shift was found between Sn(II) and Sn(IV) oxide references (edge maximum position: SnIIO, 29212.7 eV; SnIVO2, 29216.7 eV). The edge maximum position for the HCP and C-S-H samples (29215.9 to 29217.2 eV) indicates the presence of Sn(IV). This finding reveals that the formal oxidation state of Sn(IV) was not modified upon uptake by HCP or C-S-H. Reference Compounds. The EXAFS spectra (raw and calculated) for the reference compounds are presented in 2186
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 10, 2003
Figure 2. The corresponding PRDFs are shown in Figure 3. In the EXAFS spectrum for SnO2 (cassiterite) numerous and well-resolved structures appear, which are typically found in well-crystallized oxide compounds. The spectra of Sn4+(OH-) and CaSn(OH)6, despite being dominated by an oscillation with similar amplitude and frequency, are significantly different. In the case of CaSn(OH)6, a broadening of the minimum at 9 Å-1 (denoted as “A”) as well as an asymmetry of the oscillation between 9 and 11 Å-1 (denoted as “B”) are characteristic features not observed in the Sn4+(OH-) spectrum (Figure 2). In Figure 2, the k-weighted EXAFS oscillations are shown, because the typical features “A” and “B” are less visible on the k3-weighted EXAFS oscillations. The FEFF7 calculated EXAFS (dashed lines) were obtained using published crystallographic parameters for SnO2 and CaSn(OH)6 (30, 31). Figures 2 and 3 show good agreement between calculated and experimental spectra. Moreover, the typical features (“A” and “B”) of CaSn(OH)6 are correctly reproduced. Note that in the PRDF of CaSn(OH)6 some further shell peaks (“C” and “D”) appear, which correspond to 6 Ca at 4.06 Å and 12 Sn at 5.74 Å (Figure 3). In the case of Sn4+ (OH-), however, one single scattering path (6 O at 2.06 Å) is sufficient to reproduce the spectra in keeping with Sn(IV) speciation at pH 13.3, i.e., the formation of Sn(OH)62- (7). HCP and C-S-H Samples. The XAS spectra shown in Figure 2 were collected after 1 day equilibration of the samples. The dominant oscillation in the spectra of HCP and C-S-H samples is similar to the one observed in the Sn4+(OH-) and CaSn(OH)6 spectra. The typical features “A” and “B” detected in the CaSn(OH)6 spectrum were observed only in the HCP-Ov, HCP-Int, and C-S-H 1.2 samples. The spectra of HCP-Und/-Rep and C-S-H 0.7, however, neither show the typical CaSn(OH)6 features (“A” and “B”) nor are they identical to the Sn4+(OH-) spectrum. Note that the oscillation between 7 and 9 Å-1 is symmetric in Sn4+(OH-), but asymmetric in HCP-Und/-Rep and C-S-H 0.7. A clear difference between the EXAFS spectra of the HCP-Und/-Rep samples and the spectrum of the C-S-H 0.7 sample appears in the region between 5 and 7 Å-1 (wider oscillation for HCP-Und/-Rep, Figure 2). It is to be noted that a thorough interpretation of the HCP-Rep data was limited by a nonremovable glitch, which could not be eliminated because of the high signal-to-noise ratio in this sample. In Figure 4, the PRDFs of the HCP and C-S-H samples are shown in comparison to that of CaSn(OH)6. For all samples, no significant difference was observed for the first peak at 1.6 Å (nonphase shift corrected), suggesting that the octahedral oxygen coordination is maintained. For the HCP-Ov and C-S-H 1.2 samples, the PRDFs could be perfectly superimposed on the CaSn(OH)6 PRDF, including the characteristic features “C” and “D” (nonphase shift corrected R ∼ 3.5, 5.3 Å) (Figure 4a and d). This finding corroborates the similarities of the samples already observed in the EXAFS spectra. For HCP-Int, minor differences in the region 2.8-4.2 Å appear (Figure 4b). The spectrum can be reproduced by a linear combination of the EXAFS spectra of HCP-Und (20%) and CaSn(OH)6 (80%) (Figure 2). For the HCP-Und sample, a second peak (“E”) well above the noise level appears in the PRDF (nonphase shift corrected R ∼ 3.0 Å), whose position is closer to the first shell peak than peak “C” (Figure 4c). The PRDF of the C-S-H 0.7 sample is also significantly different from that of CaSn(OH)6 (Figure 4e). Note the intense second peak “F” in close proximity to the first shell peak (nonphase shift corrected R ∼ 2.8 Å). The above presentation of the PRDFs shows that the differences previously observed in the EXAFS spectra of CaSn(OH)6, HCPUnd and C-S-H 0.7 are clearly reflected in the PRDFs. Thus, CaSn(OH)6 formed in the HCP-Ov, HCP-Int, and C-S-H 1.2 samples, but was absent in the HCP-Und and C-S-H 0.7
FIGURE 3. Sn K-edge raw (___) and simulated by FEFF (- - -) modulus and imaginary part of the PRDF for reference compounds (a), HCP-Und (b), and C-S-H 0.7 (c). samples. Note, however, that the presence of a minor amount of CaSn(OH)6 (
