Environ. Sci. Technol. 2004, 38, 4423-4431
Mechanism of Europium Retention by Calcium Silicate Hydrates: An EXAFS Study M I C H E L L . S C H L E G E L , * ,† INGMAR POINTEAU,‡ NATHALIE COREAU,‡ AND PASCAL REILLER§ Laboratoire de Re´activite´ des Surfaces et Interfaces (DPC/SCP/LRSI), Laboratoire de Mesure et Mode´lisation de la Migration des Radioe´le´ments (DPC/SECR/L3MR), and Laboratoire de Spe´ciation des Radionucle´ides et des Mole´cules (DPC/SECR/LSRM), BP 11, CEA Saclay, F-91 191 Gif sur Yvette Cedex, France
The uptake of Eu by calcium silicate hydrate (C-S-H) phases as a function of Eu/sorbate ratio (from 37 to 450 µmol g-1 C-S-H), C-S-H Ca/Si mole ratio (1.3, 1.0, and 0.7), and initial supersaturating conditions was probed by solution kinetics experiments and extended X-ray absorption fine structure (EXAFS) spectroscopy, to shed light on the retention mechanism of trivalent radionuclides under waste repository conditions. The rates of Eu (9.7 × 10-10 M) uptake in C-S-H suspensions and in solutions at equilibrium with C-S-H were rapid. Uptake of more than 90% of dissolved Eu was generally observed within 15 min. Europium LIII-edge EXAFS spectra collected on samples of Eu sorbed on, or coprecipitated in, C-S-H differed from that of Eu(OH)3(s) expected to precipitate under the pH conditions of C-S-H waters, ruling out compelling precipitation of pure hydroxide phases. Fourier transforms for EXAFS spectra for Eu in sorption/coprecipitation samples displayed comparable features at distances typical of neighboring cationic shells, pointing to similar crystallochemical environments. Optimal spectral simulations were obtained by assuming the presence of Si, Si/Ca, and Ca cationic shells surrounding Eu at distances of 3.2, 3.73.8, and 3.8-3.9 Å, respectively. The nearly continuous distribution of (Si, Ca) backscattering shells parallels the distribution in Ca-(Ca, Si) interatomic distances in structural models of C-S-H. Discernible effects of experimental parameters on the Eu local environment were observed by comparison of Fourier transforms, but could not be confirmed by EXAFS quantitative analysis. These results indicate that sorbed or coprecipitated Eu is located at Ca structural sites in a C-S-H-like environment. Kinetics and spectroscopic results are consistent with either Eu diffusion within C-S-H particles or precipitation of Eu with Ca and Si creating a C-S-H-like solid phase.
* Corresponding author phone: +33 (0)1 69 08 93 84; fax: +33 (0)1 69 08 77 38; e-mail:
[email protected]. † Laboratoire de Re ´ activite´ des Surfaces et Interfaces. ‡ Laboratoire de Mesure et Mode ´ lisation de la Migration des Radioe´le´ments. § Laboratoire de Spe ´ ciation des Radionucle´ides et des Molecules. 10.1021/es0498989 CCC: $27.50 Published on Web 07/15/2004
2004 American Chemical Society
Introduction Cement materials are commonly used for conditioning of radioactive waste materials. They could be used for the construction of temporary storage facilities or as backfill material, geochemical barriers, in nuclear waste repositories. In all of these applications, concrete materials can act as physical and chemical barriers for radionuclides, for example, by favoring precipitation or sorption of important amounts of radionuclides. Therefore, they can play a decisive role in retarding the migration of heavy trace elements (1-3), including hazardous radionuclides such as Sr (4), U, Np (5), Am (6), and Cm (7). To describe the uptake of radionuclides by hardened cement pastes, macroscopic models based on thermodynamics have been developed from solution chemistry results. However, robust modeling of the interaction between radionuclides and cement materials also hinges on a correct description of the molecular mechanism of elemental uptake. The high reactivity of cement pastes can be explained by the complex mineralogy of fresh and altered cements (8). Cements are made of small-sized particles with defective structures, such as portlandite (Ca(OH)2), ettringite (Ca6Al2(SO4)3O6‚32H2O), calcium monosulfate (Ca4Al2(SO4)O4‚12H2O), and especially calcium silicate hydrates (C-S-H), the main components of cements. The defective, microcrystalline structure of C-S-H, and their high specific surface area of ca. 200 m2 g-1 (8) confer to these solid phases significant sorption affinity for cations and anions. Furthermore, C-S-H are stable under the highly alkaline conditions of the repository near-field, and therefore they may dominate the overall reactivity of solids during the early stages of cement alteration. Upon further leaching and alteration by groundwater, they undergo chemical change, resulting in a decrease of their structural Ca/Si ratio, and in structural modifications disclosed, for example, by spectroscopic techniques (ref 8 and references therein). To better understand the retention mechanism of radionuclides by these solid phases, the uptake of Cm and Eu (a chemical surrogate for trivalent actinides) by C-S-H was probed by time-resolved laser-induced fluorescence spectroscopy (TRLIFS) (7, 9). This technique can be used to discriminate structurally distinct sorption sites; however, it offers little direct insight into the crystallochemical environment of the sorbate cation. In contrast, extended X-ray absorption fine structure (EXAFS) spectroscopy can offer this desirable structural information and has been successfully used to clarify the uptake mechanism of trace elements by cement phases (5, 10-19). In the present study, the crystallochemical environment of Eu sorbed by, or coprecipitated with, C-S-H and precipitated with dissolved elements in a solution at equilibrium with C-S-H was probed by EXAFS spectroscopy. Eu was added slowly to C-S-H suspensions to mimic radionuclide feeding by leaching solutions and to minimize precipitation of hydroxide. A structural model for the Eu local environment was then developed by taking into account the C-S-H structural properties. These results allow the discussion of the mechanism(s) of Eu immobilization in C-S-H suspensions under conditions of low solution concentrations and slow sorbate addition.
Materials and Methods Solution Chemistry. All solutions were prepared from Milli-Q deionized water and reactants of American Chemical Society (ACS) or higher grade. All experiments were performed at a laboratory temperature of 22(1) °C. Calcium silicate hydrates VOL. 38, NO. 16, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4423
TABLE 1. Chemical Conditions of Eu-C-S-H Sample Preparation for EXAFS Spectroscopy [C-S-H]a (mg L-1)
Ca/Si ratio
Eu_CSH1.3 Eu_CSH1.0 Eu_CSH0.7 Eu_WCSH
500 500 500
1.3 1.0 0.7
Eu450/CSH1.3s Eu413/CSH1.0s Eu394/CSH0.7s Eu070/CSH1.0s Eu070/CSH1.0i Eu035/CSH1.3s Eu035/CSH1.0s
217 278 278 250 500 200 500
1.3 1.0 0.7 1.0 1.0 1.3 1.0
sample
a
curing time (days)
14 21 32 31 11 11 15
Eutot (µmol L-1)
µmol of Eu/ g of CSH
Coprecipitates 35.0 35.0 35.0 35.0
70 70 70
Sorption Samples 97.8 114.8 109.4 17.3 35.0 6.9 17.6
450 413 394 69 70 35 35
Eu addition (days)
18 41 9 3.6 3.6 3.6 3.6
final aging (days)
recording station
10 10 10 5
ID26 ID26 ID26 D44
31 2 22 20 10 10 6
D44 D44 D44 ID26 ID26 ID26 ID26
Amount of calcium silicate hydrate in suspension.
with Ca/Si mole ratios of 1.3, 1.0, and 0.7 (C-S-H 1.3, C-S-H 1.0, and C-S-H 0.7, respectively) were prepared by mixing fumed silica and calcium oxide (Sigma Aldrich) at desired stoichiometric ratios in argon-purged deionized water. Coprecipitate samples were obtained by adding an aliquot of an acidified (pH ) 3) 0.07 mol L-1 (M) EuCl3 solution to the suspension right after powder mixing. Coprecipitates and pure C-S-H were then cured as suspensions under constant stirring for a given amount of time (Table 1). The pH values of the equilibrium suspensions were measured with a Metrohm combination electrode and equaled 12.10(5), 11.72(5), and 9.95(5) for C-S-H 1.3, C-S-H 1.0, and C-S-H 0.7, respectively. The aqueous concentration of Ca, [Ca]aq, was measured by ionic chromatography and amounted, respectively, to 1.38(2) × 10-2, 3.6(2) × 10-3, and 1.3(2) × 10-3 M. Concentrations of dissolved silicon ([Si]aq) were measured by spectrophotometry (20) and equaled, respectively, 2(1) × 10-5, 7(1) × 10-5, and 3.86(6) × 10-3 M. These concentration values compare well with those reported for the solubility curve of C-S-H phases (8, p 146). Uptake Kinetics. Kinetics experiments of Eu uptake in C-S-H suspensions and solutions at equilibrium with C-S-H (equilibrium liquors) were performed in a CO2-free (PCO2 < 1 ppm) glovebox. C-S-H suspensions of 0.25 g L-1 were cured in polysulfone tubes for 7 days. Equilibrium liquors were then recovered by filtering suspensions at equilibrium with C-S-H (Millipore 0.22 µm nylon filter). Next, 7 mL of the suspensions, or of the equilibrium liquors, was transferred in syringes, and 100 µL of a 152Eu-labeled (2.51 × 103 Bq g-1), 6.67 × 10-8 M EuCl3 solution (CIRCA) was quickly added to obtain a total Eu concentration (Eutot) of 9.4 × 10-10 M. Note that this concentration is below the value of ∼2 × 10-9 M calculated for the solution concentration or Eu, [Eu]aq, at equilibrium with Eu(OH)3(s) at the recorded pH values (21). The syringe was capped, hand-shaken, and at given times 0.2 mL of the suspension was filtered on 0.22 µm Nylon filters (Millipore) preconditioned with Eu solutions to limit Eu loss due to adsorption on the filter. The preconditioning procedure was validated by verifying that the activity of a 152Eu solution remained unaffected upon filtration on a preconditioned filter. The [Eu]aq values were measured by mixing the filtered solution with 0.8 mL of water and 4 mL of a scintillating cocktail (Ultima Gold) and by counting for 1 h (Hewlett-Packard 2500 scintillating counter). Uncertainties on [Eu]aq were calculated at a 3σ confidence level. Sample Preparation for EXAFS Spectroscopy. Sorption experiments for EXAFS spectroscopy were conducted on great volumes (usually 4-5 L) of suspensions continuously purged with Ar to prevent CO2 contamination. Sorption was per4424
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 16, 2004
formed by introducing in the cured C-S-H suspension acidified 0.001 M EuCl3 stock solution and 0.1 M NaOH to balance the acidity of the Eu solution. The base solution was added first, and the Eu solution was introduced following either of two methods. In the first, 10-50 µL increments of the Eu stock solution were introduced under constant stirring to disperse the small aliquot of Eu solution. Periods of a few minutes between two increments were chosen to avoid [Eu]aq values exceeding the solubility limit of Eu(OH)3(s), while allowing completion of Eu addition in a reasonable delay. This method is thought to simulate the slow flux of cations at concentrations controlled by the solubility of their pure hydroxide phases. In the second, the Eu aliquot was added instantly while the suspension was stirred. Finally, the reacted Eu suspensions were allowed to equilibrate for a final period extending from 2 days to 1 month (Table 1). Sorption and coprecipitation samples were recovered by filtering the suspensions (Millipore 0.22 µm Nylon filters), slightly drying the pastes under vacuum, and pressing the pastes into pellets. Pressing usually expurgated a substantial amount of solution from the pastes, suggesting that vacuumdrying actually left much of the water present in the samples. Pellets were sealed with Kapton tapes in Teflon sample holders for XAS measurements. They were kept in boxes flushed with Ar and sealed with paraffin tape (Parafilm) prior to measurements to limit sample carbonation. Supernatant analyses for both sorption and coprecipitate samples showed that [Eu]aq values were always below the detection limit of inductively coupled plasma atomic emission spectrometry, meaning that Eu was quantitatively retained in the solid phases. In a separate experiment designed to check the formation of another solid phase, an aliquot of a 0.07 M EuCl3 solution was introduced in a solution equilibrated with C-S-H 1.3, to a final Eutot of 35 µM. A white solid was observed to form and was left to age for 5 days, and then filtered (Millipore 0.22 µm Nylon filters). The chemical composition of the solid was then obtained by dissolution in 0.1 M HNO3 and elemental analysis by inductively coupled plasma-optical emission spectrometry. This analysis and mass balance considerations yielded a bulk composition of Si0.99Ca1.87Eu4.77O11‚48H2O, where the difference between the weighted mass and the mass recalculated from chemical analysis is assumed to result from H2O molecules only. This chemical composition shows that the solid resulted from Eu coprecipitation with solution Ca and Si. The wet coprecipitate (Eu_WCSH) was loaded as a wet paste in an EXAFS sample holder sealed with Kapton windows. After collection of an Eu LIII-edge X-ray absorption spectrum, this paste was dried
TABLE 2. Structural Parameters from EXAFS Analysis of Reference Solidsa path
Rb
Nc
σ2 d (Å2)
Euaq
Eu-Og Eu-O-O Eu-O-O
2.427(4) 3.89(2) 4.46(4)
9.2(6) 48(10) 46(15)
0.0095(7) 0.0097g 0.0097g
6.5(5) 6.3g 6.3g
0.002 0.16
Eu(OH)3(s)
Eu-O1g Eu-Eu1 Eu-O-O Eu-O Eu-Eu2 Eu-Eu-O Eu-O
2.476(7) 3.65(3) 3.89(8) 3.95(5) 4.10(3) 4.35 4.45
9g 2g 36 3 6g 12 6
0.0100(4) 0.0082(38) 0.01(1) 0.01(1) 0.0082(38) 0.01(1) 0.01(1)
7.0(7) 7.0
0.007 0.025
Eu-O1g Eu-Eu1 Eu-Eu2 Eu-O2 Eu-Eu-O Eu-Eu-O
2.345(9) 3.63(2) 4.11(9) 4.19(14) 4.28(70) 4.38(70)
6 6 6 6 12 12
0.0093(6) 0.0080(30) 0.0080(30) 0.010(24) 0.010(24) 0.010(24)
5.3(1.1)
reference
Eu2O3
∆E0e (eV)
Rf f
7.0
0.008 0.021
a Number in parentheses at the end of each value indicates the uncertainties. b Interatomic distance. c Number of neighbor atoms or of distinct scattering paths for multiple scattering. d Debye-Waller factor. e Energy shift. f Residual factor Rf ) ∑k(k3χexp - k3χcalc)/∑k(k3χexp) measures the quality of the model Fourier-filtered contribution (χcalc) with respect to the experimental contribution (χexp). g First oxygen shells were fitted independently from the other shells.
TABLE 3. Quantitative Analysis of the First Eu-O EXAFS Patha ∆k rangeb
IFT rangec
Eu_CSH1.3 Eu_CSH1.0 Eu_CSH0.7 Eu_WCSH
2.4-10.3 2.4-9.7 2.4-9.7 2.9-11.3
1.5-2.5 1.5-2.5 1.5-2.5 1.5-2.5
Eu450/CSH1.3s Eu413/CSH1.0s Eu394/CSH0.7s Eu070/CSH1.0s Eu070/CSH1.0i Eu035/CSH1.3s Eu035/CSH1.0s
2.4-10.2 2.4-10.2 2.4-10.2 2.4-10.2 2.4-10.2 2.4-10.2 2.4-10.2
1.5-2.5 1.5-2.5 1.5-2.5 1.5-2.5 1.5-2.5 1.5-2.5 1.4-2.5
sample
Eu-O path REu-Od (Å)
NOe
σ2 f (Å2)
∆E0g
Coprecipitates 2.40(3) 8.8(3) 2.41(2) 9.6(5) 2.41(2) 6.8(1.4) 2.42(1) 9.1(9)
0.0066(41) 0.0091(254) 0.00102(36) 0.0118(12)
10.0(3) 10.2(4.7) 10.2(1.6) 9.5(1.0)
0.005 0.004 0.008 0.002
Sorbed 2.41(2) 2.42(1) 2.40(2) 2.40(4) 2.40(2) 2.39(2) 2.39(2)
0.0090(29) 0.0084(15) 0.0102(29) 0.0128(57) 0.0111(33) 0.0105(25) 0.0098(40)
10.6(1.4) 9.8(7) 10.3(1.3) 9.5(4.2) 9.5(1.9) 9.4(4.0) 8.5(2.9)
0.01 4 × 10-4 0.002 0.008 0.002 0.007 0.001
6.8(1.1) 6.7(6) 6.8(1.1) 11(5) 9(2) 9(4) 8.6(2.8)
Rfh
a Number in parentheses at the end of each value indicates the uncertainties. b Fourier-transformed χ(k) range. c Range of the inverse FT. Interatomic distance. e Number of neighbor oxygens. f Debye-Waller factor. g Energy shift. h The residual factor Rf ) ∑k(k3χexp - k3χcalc)/∑k(k3χexp) measures the quality of the model Fourier-filtered contribution (χcalc) with respect to the experimental contribution (χexp). d
slowly and probed by X-ray diffraction on a Siemens D500 diffractometer (Co KR radiation). Only a wide band centered at 3.1 Å was observed, suggesting the absence of long-range crystallographic ordering in the Eu_WCSH coprecipitate. Eu2O3 and Eu(OH)3(s) references (Table 2) were prepared by diluting the solids in boron nitride to achieve an X-ray absorption edge jump of ∼1, and then pressing the preparations. Solvated Euaq was obtained by dissolving Eu2O3 in 0.4 M HClO4 to a final Eu concentration of 0.1 M. The solution was then loaded in a Teflon cell sealed by Kapton windows. EXAFS Spectroscopy. EXAFS spectra for Euaq, Eu2O3, Eu(OH)3(s), for high concentration samples (i.e., Eu450/ CSH1.3, Eu413/CSH1.0s, and Eu394/CSH0.7s), and for Eu_WCSH were collected at the D44 beamline of the Laboratoire pour l’utilisation du rayonnement e´lectromagne´tique (LURE; Orsay, France). The storage ring was operated at an energy of 1.85 GeV and a current of ∼300 mA. The beam energy was selected with a Si(111) double crystal monochromator, which was detuned by 30% of the beam maximal intensity to remove higher order harmonics. The intensity of the incident beam was measured using an airfilled ionization chamber. Spectra for Euaq, Eu2O3, Eu(OH)3(s), and Eu_WCSH were measured in transmission mode using air-filled ionization chambers. Spectra for high con-
centration samples were collected in fluorescence mode using a multi-element solid-state Ge detector (Canberra). High dilutions samples were recorded on the ID26 beamline at the ESRF (Grenoble, France), with a storage ring energy of 6 GeV and a ring current between 200 and 100 mA. The optics of the beamline consist of a flat water-cooled Cr mirror for higher harmonics rejection, a Si(111) double crystal monochromator cooled to ∼133 K, and a pair of beamfocusing, higher harmonics-rejecting mirrors (Si and fused silica). The intensity of the incident beam was quantified by using diode array detectors. The fluorescence signal was measured by using a Cr filter and diode detectors. For each sample, several scans were collected in Quick EXAFS mode, then compared, and rejected if too noisy, and the selected scans were summed. Analysis of EXAFS data was performed following standard procedures by using Athena and Artemis interfaces to the Iffefit software (22) (Table 3). Atomic absorption was removed, and the EXAFS spectra were extracted by using the Autobk routine (23). The Autobk algorithm extracts the EXAFS signal with a spline that minimizes the low-R portion of the Fourier transform (FT) that is not directly related to the presence of backscattering shells. The EXAFS spectra were apodized with a Kaizer-Bessel window (weight ) 2) and were FourierVOL. 38, NO. 16, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4425
implemented in FEFFIT. This uncertainty does account for the accuracy of fitted values, which can be guessed only by comparing the values obtained by EXAFS spectroscopy and by other, more accurate techniques.
Results and Interpretation
FIGURE 1. Kinetics of Eu uptake in calcium silicate hydrates (C-S-H) suspensions and in solutions at equilibrium with C-S-H (“equilibrium liquors”). Black symbols: Eu uptake in 0.75 g L-1 C-S-H 1.3 ([), C-S-H 1.0 (b), and C-S-H 0.7 (2) suspensions. White symbols: Eu uptake in C-S-H 1.3 (]), C-S-H 1.0 (O), and C-S-H 0.7 (4) equilibrium liquors. Lines guide the eye. transformed. The FT modules display amplitude maxima or peaks at apparent distances (R + ∆R), which differ from half the length of the structural scattering path (R, equal to the absorber-backscatterer distance for simple scattering) by ∼ -0.3 Å, because of phase shifts of the EXAFS waves (24). Selected FT peaks were Fourier back-transformed and fitted by using phases and amplitude functions calculated with FEFF7 (25), and crystallographic data for Eu(OH)3(s) (26), Eu2O3 (27), and 1.1 nm tobermorite (28, 29) in which one Ca was replaced by Eu. The amplitude reduction factor (S02) was set to 1.0 to correctly reproduce the number of neighboring atoms in the structural reference. Note that S02 ) 1.0 agrees with earlier studies of lanthanide coordination in water (30). The oxygen shell of the first coordination sphere was fitted first, and then the next-nearest atomic shells. Finally, both fits were optimized to yield an optimal solution. For all samples, the statistical uncertainty resulting from spectral noise was calculated by using the procedure
Solution Chemistry. Kinetics experiments show that, immediately after Eu introduction in solution, [Eu]aq values decrease rapidly in both C-S-H suspensions and equilibrium liquors (Figure 1). More than 90% of dissolved Eu was removed from the solution within 15 min, except for the liquor at equilibrium with C-S-H 0.7. This rapid removal indicates a quantitative Eu uptake either on solid phases or in solution (co)precipitates. Further data analysis reveals that, except for C-S-H 1.3, [Eu]aq values in C-S-H suspensions are always lower than those observed in equilibrium liquors after 12 min of reaction. A tentative explanation may be that Eu removal in both suspension and equilibrium liquor results from coprecipitation in solution with species release by C-S-H dissolution. Such coprecipitation would deplete the supernatant from dissolved species, but this depletion can be balanced by C-S-H dissolution when C-S-H are present. However, given the difference in concentration of dissolved Eu (10-5 M), any Eu coprecipitate is unlikely to appreciably affect [Ca]aq and [Si]aq, even in the equilibrium liquors. Therefore, Eu uptake in C-S-H 1.0 and C-S-H 0.7 must result, at least in part, from Eu sorption on the C-S-H. Note that such a sorption mechanism is usually limited by the amount of available sorption sites. In contrast, [Eu]aq values in the C-S-H 1.3 suspension and C-S-H 1.3 equilibrium liquor are similar, suggesting that a significant, if not predominant, fraction of Eu in the suspension (co)precipitates. EXAFS Spectroscopy of Reference Compounds. The EXAFS spectrum for Euaq displays only a single wave frequency of monotonically decreasing amplitude for k > 3 Å (Figure 2a). This spectrum is consistent with the presence of a single ordered coordination sphere. In contrast, EXAFS
FIGURE 2. (a) k3-weighted Eu LIII-edge EXAFS spectra for Eu in solution (Euaq), Eu(OH)3(s), and Eu2O3. (b) Eu LIII-edge experimental (solid line) and simulated (dotted lines) moduli and imaginary parts of the Fourier transforms for reference compounds. 4426
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 16, 2004
FIGURE 3. (a) Eu LIII-edge EXAFS spectra for Eu coprecipitated with Ca in C-S-H and for Eu sorbed on C-S-H. Note that spectra collected at ID26 all display a nonremovable glitch at 10 Å-1. (b) FT magnitudes and imaginary parts for Eu coprecipitated in C-S-H with Ca/Si ratios of 1.3 (Eu_CSH1.3), 1.0 (Eu_CSH1.0), and 0.7 (Eu_CSH0.7). (c) FT magnitudes and imaginary parts for Eu_CSH1.0 and Eu adsorbed by C-S-H 1.0. Eu was added either slowly (Eu070/CSH1.0s) or instantly (Eu070/CSH1.0i). (d) FT magnitudes and imaginary parts for Eu uptake by C-S-H 1.0 at increasing surface coverage. See Table 1 for more details on experimental parameters. spectra for both Eu(OH)3(s) and Eu2O3 have distinct frequencies, which are absent from Euaq, and thus cannot be attributed only to multiple scattering (MS) paths with the first coordination sphere. Therefore, they can be related to the presence of higher atomic shells. Fourier transforms all display a first peak near 2.0 Å, that can be related to contributions from oxygens of the nearest coordination sphere (Figure 2b). High amplitude contributions at R + ∆R > 2.5 Å can also be observed for Eu(OH)3(s) and Eu2O3 and originate mainly from next-nearest Eu scattering shells. In contrast, only weak contributions at ∼3.2 and 3.9 Å were observed on the Euaq FT. They may originate either from MS paths within the first coordination sphere or from single scattering paths from second and more distant hydration spheres. Figure 2b shows that the spectrum for Euaq could be correctly modeled (R ) 0.002) by assuming a number of oxygens NO ) 9.2(6) at a Eu-O distance REu-O ) 2.427(4) Å (σ2 ) 0.0095(7) Å2; Table 1). The reported distances and number of neighbor atoms compare with the values reported in a previous study (30), but they slightly differ from REu-O ) 2.45 Å and NO ) 8.4 obtained from X-ray diffraction (XRD) measurements in solutions (31). The slight discrepancy between EXAFS- and XRD-derived REu-O and NO may stem from the accuracy of the respective techniques. Higherdistance contributions at R + ∆R > 2.5 Å were correctly modeled by assuming two triangular MS paths between Eu and pairs of oxygens from the first shell (Eu-O-O paths), at REu-O-O1 ) 3.89 Å and REu-O-O2 ) 4.46 Å (NOO1 ) 48(10)
and NOO2 ) 46(15); sOO12 ) sOO22 ) sO12). No attempt was made to model these contributions by assuming the presence of a distant hydration sphere. Good structural models were also obtained for Eu(OH)3(s) and Eu2O3, with distances differing only marginally from crystallographic data (Table 1). Neglecting multiple scattering paths lead to a substantial decrease of the fit quality factor Rf for both Eu(OH)3(s) (Rf ) 0.025 and 0.16 with and without multiple scattering, respectively) and Eu2O3 (Rf ) 0.021 and 0.042 with and without multiple scattering, respectively). EXAFS Spectroscopy of Coprecipitation and Sorption Samples. For all sorption and coprecipitation samples, an intense absorption line at 6883 eV (data not shown) dominates the X-ray absorption edge. The position of this line compares with that observed for trivalent Eu in solids (∼6883-6885 eV) and is located at significantly higher energy than the maximum of X-ray absorption for divalent Eu (∼6973 eV) (32). This shows that Eu was trivalent in all of our sorption/ coprecipitation samples. EXAFS spectra obtained for Eu coprecipitated with C-S-H look similar to each other, pointing to comparable crystallochemical environments for Eu, regardless of the preparation protocol (Figure 3a). These spectra somewhat differ from those obtained for the reference compounds (Figure 2). For example, the oscillations at 8 Å-1 for the Eu coprecipitates are broadened with respect to that for Euaq. Spectral differences between Eu coprecipitates and Eu(OH)3(s) are even more obvious (e.g., near 7.5 Å-1), ruling out compelling formation of Eu(OH)3(s) upon coprecipitaVOL. 38, NO. 16, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4427
tion. This result corroborates earlier findings based on TRLIFS (32). In contrast, no significant dissimilarity could be observed between EXAFS spectra for Eu coprecipitated with, or sorbed by, C-S-H (Figure 3a). FT amplitudes and imaginary parts for all samples display a major peak at R + ∆R ) 2.0 Å, which can be related to the contribution of nearest O shells at ∼2.4 Å (Figure 3b-d). Several FT peaks at R + ∆R values of 2.9, 3.2, and 3.7 Å can also be observed for all Eu coprecipitated with C-S-H (Figure 3b). The FTs compare well in positions of amplitude maxima and imaginary parts, again pointing to comparable molecular environments. These maxima have greater amplitudes than for Euaq, meaning that they can result only from EXAFS contributions of next-nearest cationic shells. Figure 3c reveals that the FT for Eu coprecipitated with C-S-H (Eu_CSH1.0), and for Eu added either incrementally (Eu070/CSH1.0s) or instantly (Eu070/CSH1.0i) in a C-S-H suspension differ slightly by the position and amplitude of the FT envelope near 2.8 and 3.3 Å. Specifically, the FT maximum near 2.8 Å displays greater amplitude for Eu_CSH1.0 than for the other samples. However, imaginary parts are roughly comparable; that is, the oscillation minima of imaginary parts at R + ∆R ) 3.2 Å coincide with amplitude maxima. The relative dephasing of the FT imaginary part and amplitude maxima can be used to disclose changes in nature of predominant backscatterers (see, e.g., 33). The absence of strong variations in relative positions of FT peaks and imaginary parts therefore points to similar backscattering cations. Finally, Figure 3d shows that increasing Eu uptake results in minor variations in the FT contributions at R + ∆R > 2 Å. For example, the maxima in amplitude and imaginary part for Eu070/CSH1.0s are located at R + ∆R ) 3.35 and 3.5 Å, respectively, and both extrema are shifted toward lower R + ∆R values (to 3.3 and 3.4 Å, respectively) for Eu035/CSH1.0s. In contrast, the maximum in imaginary part at R + ∆R ≈ 3.4 Å for Eu_WCSH coincides in position with that of Eu035/CSH1.0s, but the amplitude maximum for the coprecipitate is located near 3.15 Å. Therefore, Eu environments may be roughly comparable at high and low coverage, but they differ from that of Eu in Eu_WCSH (Figure 3d). The FT peak at 2.0 Å was modeled for all samples by assuming the presence of a single oxygen shell. The fits yielded NO values between 6.8(1.4) and 11(5), and REu-O distances between 2.39(2) and 2.42(1) Å. The REu-O distances are longer than REu-O ) 2.35 Å for hexacoordinated Eu in Eu2O3, and marginally shorter than REu-O ) 2.43 Å for octocoordinated Eu in water. This suggests that sorbed Eu is seven-fold to eight-fold coordinated. Interestingly, REu-O ) 2.41 Å compares well with average Ca-O distances for heptacoordinated Ca (〈RCa-O〉 ≈ 2.46 Å) as in 1.1 nm tobermorite, an ordered phase related to C-S-H gels (28, 29). The similarities in the cationoxygen distances and coordination numbers agree with possible substitution of Ca by Eu at Ca structural sites. Quantitative analysis of the higher cationic shells was further performed by assuming that Eu substituted for Ca in C-S-H gels, and by noting that such gels are structurally similar to jennite and 1.4 nm tobermorite (8). The crystal structure of these mineral phases has not been fully resolved yet, but has been proposed to resemble the layered structure of 1.1 nm tobermorite (28, 29). Tobermorite layers are made by a central CaO2 sheet, which is bound on both sides by silicate chains. Every third tetrahedron of these chains points outward the layers and is named, in the cement terminology, a “bridging” tetrahedron. The nature of interlayer binding is disputed. On one hand, Hamid (28) suggested that tobermorite layers were bonded together by interlayer partially hydrated Ca. In the proposed crystallographic structure, however, interlayer Ca would be surrounded by a single oxygen at 3.35 Å, and then by six oxygens at unrealistically 4428
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 16, 2004
FIGURE 4. Experimental (solid lines) and modeled (dotted lines) Fourier-filtered k3χ(k) contributions for the next-nearest backscattering shells at R + ∆R distances spanning the [2.5, 4.0 Å] interval. high distances spanning the [2.93, 3.32 Å] interval. This casts some doubt on the accuracy of this Ca atomic position. On the other hand, Merlino et al. (29) suggested that bridging tetrahedra shared their apexes across the interlayer space. In this structure, Ca located in CaO2 sheets are surrounded by about the same numbers of neighboring cations within 4 Å, and at about the same interatomic distances, as in the Hamid structure (28). However, bridging tetrahedra would be connected each to three adjacent Si tetrahedra. This is in contradiction to results from nuclear magnetic resonance (NMR) showing that Si tetrahedra in C-S-H and C-S-Hlike 1.1 nm tobermorite are bridged to only one or two adjacent Si groups (in NMR terminology, Q1 and Q2 tetrahedra, respectively; 34). We propose that the structural organization of the C-S-H interlayer resembles that of the 1.1 nm tobermorite structure described by Hamid (28), but we believe that interatomic distances for Ca are more reasonable, if more accurate, in the work of Merlino et al. (29). In this structure, Ca from the central CaO2 sheet is surrounded, on average, by seven oxygens at an average Ca-O distance 〈RCa-O〉 ) 3.46 Å, by 6 Si at RCa-Si distances from 3.07 to 3.68 Å (average 〈RCa-Si〉 ) 3.47 Å), and by 6 Ca at RCa-Ca distances from 3.67 to 3.97 Å (average 〈RCa-Ca〉 ) 3.83 Å). Some Ca are also located in the interlayer, where they are
a Fourier-transformed χ(k) range. b Range of the inverse FT. c Energy shift. d The residual factor R ) ∑ (k3χ 3 3 f k exp - k χcalc)/∑k(k χexp) measures the quality of the model Fourier-filtered contribution (χcalc) with respect to the experimental contribution (χexp). e Values coupled during the fitting procedure.
0.022 0.004 0.022 0.027 0.023 0.028 0.024 10.6 9.8 10.3 10.9 9.5 10.2 9.9 0.0046(4) 0.0026(18) 0.0025(32) 0.0049(14) 0.0046 0.0045(12) 0.0056(70) 4.1(9) 2.2(6) 2.2(9) 2.9(1.5) 1.2(1.5) 4.7(4.0) 2.7(2.2) 3.78(3) 3.80(4) 3.76(7) 3.84(7) 3.81(14) 3.82(10) 3.81(10) 4.2(1.1) 4.0(8) 4.4(1.7) 3.4(1.8) 4(3) 5.2(4.8) 4(3) 2.4-10.2 2.4-10.2 2.4-10.2 2.4-10.3 2.4-10.2 2.4-10.3 2.4-10.2 Eu450/CSH1.3s Eu413/CSH1.0s Eu394/CSH0.7s Eu070/CSH1.0s Eu070/CSH1.0i Eu035/CSH1.3s Eu035/CSH1.0s
2.5-4.1 2.5-4.0 2.5-4.0 2.5-4.0 2.5-4.0 2.5-4.0 2.5-3.9
3.23(6) 3.19(12) 3.21(11) 3.13(5) 3.13(7) 3.06(26) 3.08(27)
0.6(4) 0.2(3) 0.6(2) 0.6(5) 0.6(6) 0.3(1.3) 0.2(8)
0.0025(25)e 0.0026(25)e 0.0044(16) 0.0025(46)e 0.003(10)e 0.0037(29)e 0.0051(66)e
Sorption Samples 3.70(3) 3.75(3) 3.70(4) 3.78(6) 3.74(10) 3.76(11) 3.75(10)
0.0025(25)e 0.0026(25)e 0.0065(16) 0.0025(46)e 0.003(10)e 0.0037(29)e 0.0051(66)e
0.028 0.029 0.017 10 9.8 10.3 0.0047(36) 0.0044(74) 0.0037(23) 2(4) 1.1(1.1) 0.4(5.0) 3.89(18) 3.87(37) 3.80(57) 4.2(2.7) 2.9(1.8) 3.1(5.3) 2.4-10.3 2.4-9.7 2.4-10.2 Eu_CSH1.3 Eu_CSH1.0 Eu_CSH0.7
2.5-4.0 2.5-4.0 2.5-4
3.17(5) 3.17(6) 3.17(3)
1.8(1.3) 1.4(1.1) 2.0(1.1)
0.0025(16)e 0.0024(13)e 0.0026(23)e
Copreci pitates 3.81(11) 3.80(16) 3.75(28)
0.0025(16)e 0.0024(13)e 0.0026(23)e
Rf d ∆E0c σ2 (Å2)
NCa
Eu-Ca shell
REu-Ca (Å) σ2 (Å2)
NSi
Eu-Si shell
REu-Sit(Å) σ2 (Å2) Eu-Si shell
NSi REu-Si (Å) IFT rangeb ∆ka sample
TABLE 4. Quantitative Analysis of Higher EXAFS Paths
surrounded by 3 Si at RCa-Si distances from 3.20 to 3.68 Å (average 〈RCa-Si〉 ) 3.37 Å). A nearly continuous distribution in of Eu-Si and Eu-Ca distances would thus be obtained upon Eu-for-Ca substitution, complicating the fitting procedure. However, EXAFS is only slightly sensitive to the atomic number Z of the backscatter. Because ZSi ) 14 for Si and ZCa ) 20 for Ca, ZSi and ZCa do not differ too much, and a mixed (Si, Ca) contribution would be correctly modeled by assuming a Si shell only. On the basis of this inference, correct fits for the coprecipitates samples were obtained (Figure 4) by assuming two Si backscattering shells at radial distances REu-Si ≈ 3.1 and 3.7 Å, and one Ca shell at radial distances REu-Ca ≈ 3.8-3.9 Å (Table 4.). Note that the EXAFS contributions of these shells overlap and such overlaps can cause a significant increase in the uncertainty in EXAFS-derived structural parameters (35). Therefore, although EXAFSderived numbers of neighbor Si, NSi, at ∼3.2 Å appear to decrease for the sorption samples relative to the coprecipitates, the too high uncertainty on NSi hampers any definite conclusion. Finally, although contributions from MS paths within the first oxygen shell could not be excluded, they were not required to improve the fit. Indeed, only weak MS contributions from the nearest oxygen shell would be expected if this shell were distorted, as is the case for Ca in the C-S-H structure. In Eu_WCSH, Eu is 5 and 3 times as abundant as Si and Ca and therefore is probably surrounded by Eu, Ca, and possibly Si shells. The presence of neighboring Eu shells is further supported by the spectral differences observed between Eu_WCSH and the other samples. Hence, the nextnearest EXAFS contributions were correctly modeled (Rf ) 0.018) by assuming 0.4(2) Si at 3.19(3) Å (σ2 ) 4.2 × 10-3 Å2), 0.4(2) Ca at 3.65(3) Å (σ2 ) 9.4 × 10-3 Å2), and 0.6(4) Eu at 3.90(5) Å (σ2 ) -9.6 × 10-3 Å-1; ∆E0 ) 9.5 eV and Rf ) 0.019). These interatomic distances fall within the range of Eu-(Si, Ca) and Eu-Eu distances for Eu structural references, and for Eu coprecipitated with C-S-H phases.
Discussion Analysis of EXAFS spectra for the C-S-H samples indicated that little or no precipitation of the Eu(OH)3(s) phase has occurred, despite Eutot concentrations well above supersaturating conditions. This result confirms previous evidence from TRLIFS for the absence of Eu(OH)3(s). Also, formation of fully hydrated outer-sphere complexes only can be ruled out because of the differences between the spectra of Euaq and sorption/coprecipitation samples. Therefore, Eu uptake in C-S-H suspensions may be explained only by Eu sorption at the C-S-H surface, diffusion and possibly Ca substitution within the structure, or precipitation of a mixed hydroxide with species released by C-S-H dissolution. Structural Models for Eu Coprecipitated with C-S-H. Previous studies have suggested that Eu can substitute for Ca in the C-S-H structure (7, 36). This inference is based on the similarity in ionic radii of Eu and Ca in seven-fold coordination (1.01 and 1.06 Å, respectively) (37). Further insight into the Eu distribution between layer and interlayer Ca sites might be sought from the number of neighboring cations. However, such a formal distribution is hindered by the high uncertainty in the EXAFS-derived number of neighboring cations. For example, the presence of a Ca shell at REu-Ca ) 3.8-3.9 Å clearly points to Eu located in CaO2 sheets, but NCa values for this shell are affected by a too large uncertainty to be quantitatively meaningful. Also, the total number of Ca and Si neighbors (5.4 e NCa,Si e 8) detected by EXAFS are always lower that the number expected for Eu in CaO2 sheets of tobermorite layers, and greater that NSi ) 3 for interlayer Ca. However, NMR studies of C-S-H showed the ratio of Q1 over Q2 tetrahedra increases, and thus the length of Si chains decreases, with increasing Ca/Si ratio VOL. 38, NO. 16, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4429
FIGURE 5. Structural model of Eu sorption sites on C-S-H structure. C-S-H layers are modeled by a CaO2 sheet ligated on both sides by short silicate chains. Ca sites are labeled as in the 1.1 nm tobermorite structure of Hamid (28). Eu can be located either at a Ca structural site within the solid structure or can adsorb at the C-S-H surface. (38). This chain shortening can be incorporated in the tobermorite structural model by assuming omission of bridging tetrahedra, resulting in a decrease in the number of Si surrounding the layer Ca with increasing Ca/Si. Omission of Si tetrahedra can also result in structural distortion, increasing the structural disorder around Eu. This increase in structural disorder can result in an apparent decrease in number of neighboring atoms, the magnitude of which cannot be fully comprehended. In conclusion, EXAFS results are consistent with Eu substituting for Ca in one or several Ca sites of a tobermorite layerlike C-S-H structure, and the precise distribution of Eu in these sites cannot be confidently determined yet. Structural Models for Eu Sorbed on C-S-H. The spectral similarities observed for coprecipitation and sorption samples indicate that Eu sorbed by C-S-H are located at Ca structural sites of a C-S-H structure (Figure 5). These sites can be located either at the C-S-H surface, thus allowing surface complexes, or within the bulk solid, filled by Eu as a result of either diffusion or coprecipitation. Eu forming surface complexes would be surrounded by a significantly smaller number of neighboring cations than in the bulk structure. This hypothesis appears at variance with the comparable EXAFS-derived numbers of neighboring cations observed for sorption and coprecipitation samples. Therefore, a significant fraction of sorbed Eu is actually located within solid phases, as a result of solid diffusion or coprecipitation. These two uptake mechanisms may be discriminated, for example, by following solution parameters such as pH and elemental concentrations of dissolved species during sorption. If coprecipitates form in significant amounts, they should alter 4430
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 16, 2004
the concentrations of dissolved species. However, in most of our sorption experiments, only a small [Eu]aq (
