Aquat Geochem (2009) 15:443–456 DOI 10.1007/s10498-009-9060-9 ORIGINAL PAPER
Uranyl Retention on Quartz—New Experimental Data and Blind Prediction Using an Existing Surface Complexation Model Florian Huber Æ Johannes Lu¨tzenkirchen
Received: 3 June 2008 / Accepted: 1 January 2009 / Published online: 27 January 2009 Ó Springer Science+Business Media B.V. 2009
Abstract The adsorption behaviour of uranyl onto seven different samples of quartz was studied in batch experiments. Sea-sand (0.1–0.3 mm), Fil-Pro 12/20 (1–2 mm) and five Min-U-Sil samples with smaller particle sizes (5, 10, 15, 30 and 40 lm) were used. The uptake curves show ‘‘pH adsorption edges’’ in the range of pH 4–5. A good agreement of the new data with literature data was found when plotting surface-normalised distribution coefficients versus pH. Differences in the adsorption behaviour for pre-treated and untreated sea-sand samples were detectable resulting in a shift of the pH edge to higher pH values after treatment. A literature surface complexation model was applied for blind predictions of the experimental results. The simulations described the experimental observations quite well for the Min-U-Sil samples. For the two coarser quartz samples, the calculated over-predictions were explained by the larger-than-expected measured specific surface area and measurable amounts of associated minerals, for Fil-Pro 12/20 and seasand, respectively. Dissolution of the samples was studied as a function of pH. After 5 days, the measured Si concentrations were all higher than equilibrium quartz solubilities, but lower than those of amorphous silica. With increasing pH, dissolved silica increased. This strongly suggests that formation of dissolved uranyl–silicato complexes have to be considered based on measured silica concentrations. Keywords Surface complexation model Adsorption Batch experiments Uranium (VI)
1 Introduction The release of uranium into the environment through anthropogenic activities (e.g. migration from a nuclear waste repository or from mine waste) is a potential threat to ecoand biosphere due to uranium chemo- and radiotoxicity. Various processes occurring in the F. Huber (&) J. Lu¨tzenkirchen Forschungszentrum Karlsruhe (FZK), Institut fu¨r Nukleare Entsorgung (INE), Karlsruhe Institute of Technology (KIT), P.O. Box 3640, 76021 Karlsruhe, Germany e-mail:
[email protected]
123
444
Aquat Geochem (2009) 15:443–456
subsurface environment influence the migration of uranium leading to retardation e.g. through the adsorption onto matrix material or the precipitation of UO2 under anoxic conditions caused by abiotically or microbially induced redox processes. Attachment of dissolved uranium to colloids, which can be transported rapidly over relatively long distances, may enhance the mobility, although to a minor extent compared to other radionuclides like e.g. Pu. Hence, it is of particular importance to gain insight into the processes controlling the mobility and migration of potential contaminants such as uranium for the development of sound risk-assessment strategies or high-level nuclear waste disposal concepts. In this study, we are interested in the behaviour of U(VI), i.e. oxidising conditions. The linear uranyl (UO22?) is the thermodynamically stable form of uranium under oxidising conditions (Grenthe et al. 1992), which forms strong aqueous complexes with different ligands. Several studies about the adsorption of uranium(VI) from aqueous solutions onto a wide range of solid materials (e.g. Fe- and Al-oxides/hydroxides, clay and quartz) under different chemical conditions (pH, ionic strength, presence of different complexing ligands, total quantity of uranium, etc) are available. All the studies involving silica or quartz revealed a more or less strong impact of the solution composition on the adsorption of uranyl (Kohler et al. 1996; Jung et al. 1999; Artinger et al. 2002; Hongxia and Zuyi 2002; Davis 2001; Pathak and Choppin 2006; Mibus et al. 2007; Gabriel et al. 2001; Zuyi et al. 2000; Pathak and Choppin 2007a, b; Zuyi et al. 2004). Because of the relatively easy procedure, batch experiments are the common way to study the adsorption behaviour of e.g. heavy metals. Upscaling of batch results to contaminant transport in the environment should be done carefully due to the static nature of the batch technique and the dynamic nature of the migration process. This concerns, for example, the particle size: in laboratory experiments most frequently colloidal materials are used, whereas matrix material such as quartz in the environment usually has much larger grain sizes. In this respect, more realistic approaches to study the behaviour of contaminants in contact with potential sorbents are (dynamic) column experiments. These, however, require more effort than batch experiments and are more prone to experimental as well as interpretational problems. Results of batch experiments might be considered as first information about a given system which in a second step should be transferred to more realistic conditions by carrying out the respective column experiments. Another frequently encountered problem with batch experiments is that in most cases a full analysis of the aqueous ‘‘equilibrium’’ solution is not available. Thus, for the quartz system studied here, it is important to take into account sorbent dissolution in order to be able to evaluate the potential formation of aqueous uranyl–silicato complexes. This phenomenon has previously been largely disregarded. Whenever considered in calculations, the assumption of thermodynamic equilibrium is made, implying that quartz or silica would show equilibrium solubilities. The aims of this study were: (1) (2)
(3)
experimental determination of adsorption data from batch experiments for the system uranyl–quartz in CO2 free systems: such data are missing for the Min-U-Sil samples, experimental determination of adsorption data from batch experiments for a range of quartz samples: samples were to span a range of physical properties and the results were to be compared to available literature data, comparison of the experimental results with blind predictions using a model calibrated on batch and column data (in the presence of carbonate, but under those conditions where carbonate is not expected to affect results significantly): this is a test
123
Aquat Geochem (2009) 15:443–456
(4)
445
for an existing model on independent data and different samples of a nominally identical sorbent, evaluation of the impact of sorbent dissolution on the speciation of uranyl in solution.
2 Materials and Methods 2.1 Materials Seven different quartz samples were used: Min-U-Sil 5(lm), 10, 15, 30, 40 (referred to as MUS hereafter), Fil-Pro 12/20 (*1 mm grain size; referred to as FP) and sea-sand (*0.1– 0.3 mm; referred to as SS). SS was obtained from Merck, while all other samples were kindly provided by US Silica Company. To guarantee reproducible quartz samples, the materials were pre-treated to remove impurities. First the quartz samples were heated to 600°C for 48 h to remove organic material, afterwards the samples were washed with 4 M HNO3 to remove Fe- and Al-oxides/hydroxides and finally washed at least five times with ultrapure deionised water (Milli-Q) to remove the acid. Solutions and suspensions were prepared with ultrapure deionised water. Solutions and suspensions were either acidified or bubbled with purified Argon (see below) before use, whereby the impact of CO32- is minimised. As background electrolyte, NaNO3 (Merck suprapur) was chosen and adjusted to an ionic strength of 0.01 M to be compatible with the experimental work of Kohler et al. (1996). The original uranyl solution was an acidic (HNO3) 1,000 ppm ICP-MS standard solution, from which U(VI) stock solutions were prepared. Aliquots were taken from these stock solutions to check the U(VI) concentrations in these stock solutions via ICP-MS at regular intervals in the course of this study. 2.2 Solid Characterisation The solid samples were characterised by X-ray photoelectron spectroscopy (XPS) before the pre-treatment, and by energy disperse X-ray microanalysis (EDX) and scanning electron microscopy (SEM) after pre-treatment. According to XPS analysis, all the MUS samples can be considered as relatively pure containing [99% quartz with some minor amounts of C, Al and Fe. XPS results for FP showed amounts of C (*6 at.%), Fe (*0.4 at.%), Al (0.7 at.%) and Ti (0.1 at.%). Regarding SS specimens, relatively high values of Al (*5.7 at.%) and K (*1 at.%) were detected, which was later corroborated through EDX analysis and interpreted as small amounts of feldspar. The relatively high amount of impurities on the seasand motivated a separate study of uranyl adsorption before and after treatment. EDX analysis results showed that with the exception of SS the different samples after pretreatment are very pure consisting only of Si and O with a ratio of 1:2. The pre-treated SS still contained minor amounts of Al and K. Specific surface areas of the different particles were determined by BET N2-adsorption. The samples were heated to 300°C and degassed. The determination of the surface area was done via the multiple-point-method at different pressures and a subsequent fit with the BET isotherm. Table 1 lists the specific surface areas for all the used quartz samples together with their grain size, brand name and abbreviations. Kohler et al. (1996) reported a lower specific surface area of 0.33 m2/g for MUS 30. Vuceta (1976) reported a specific
123
446
Aquat Geochem (2009) 15:443–456
Table 1 Sample names, abbreviations used in the text, grain sizes and specific surface areas of the used quartz samples Brand name
Abbreviation
Grain size (lm)
Specific surface area (m2/g)
Min-U-Sil 5
MUS5
5
6.5
Min-U-Sil10
MUS10
10
3.6
Min-U-Sil15
MUS15
15
2.4
Min-U-Sil30
MUS30
30
2.1
Min-U-Sil40
MUS40
40
1.8
Sea-sand (pre-treated)
SS
100–300
0.93
Sea-sand (un-treated)
SS
100–300
0.69
Fil-Pro 12/20
FP
1,000–2,000
0.24
surface area of 4.2 m2/g for MUS 5. The comparatively high surface areas for our MUS samples are likely due to the presence of finer particles which were observed by SEM. In general, all the MUS samples are very poorly sorted and contained fines. Note that Kohler et al. (1996) removed the fines prior to their experimental work. The two coarser quartz samples also exhibit relatively high surface areas based on the nominal grain size. In contrast to the MUS samples, both SS and FP have no noticeable amounts of fines. In the case of FP (0.24 m2/g), the ‘‘high’’ surface area is likely due to pores and heterogeneities on the surface of the particles which can be seen in the SEM pictures (Lu¨tzenkirchen and Huber 2007). To characterise the micro-porosity of the FP sample, a complete gas adsorption isotherm was measured. The observed hysteresis is typical for micro-porous materials. From the full adsorption isotherm, an additional value for the specific surface area was calculated resulting in a slightly higher value compared to the standard-BET measurement (0.33 vs. 0.24 m2/g). Specific surface area measurement for pre-treated SS yielded a value of 0.93 m2/g and a slightly lower value of 0.69 m2/g for untreated SS. In all the BET-measurements, care was taken to have sufficient total surface for significant measurements. Pokrovsky et al. (2006) have shown that it is actually possible to measure very low surface areas of nearly 1-mm grain size quartz samples. 2.3 Experimental Procedure All the experiments were done at ambient pressure and room temperature (21°C ± 1). To minimise the impact of CO2, the samples were bubbled with Ar which was previously flushed two washing bottles (NaOH and 0.01 M NaNO3, respectively). To obtain comparable results for the different batch samples, defined amounts of the different solids were weighed depending on their measured specific surface area, and a defined amount of uranyl was added to obtain identical uranyl to surface ratios (0.63 ± 0.1 nm/m2). For Min-U-Sil 40 a value half of that was used. For the sea-sand sample a variation of the ratio was studied. In all cases, high solid-solution ratios were employed to minimise wall effects. In addition, blanks were analysed for wall effects but no measureable U(VI) retention in the test tubes was detected over the whole pH range studied for the relevant total U(VI) concentrations. All the experiments were carried out in the following way: known masses of quartz samples were delivered into test tubes (Sarstedt, 15 ml tubes), and required volumes of
123
Aquat Geochem (2009) 15:443–456
447
Milli-Q water (bubbled with purified Ar), 1 M NaNO3 solution (to obtain 0.01 M), 0.01 M HNO3 solution and defined volumes of the U(VI) stock solution were added. This resulted in a suspension at low pH where U(VI) was not adsorbed. The last step was the adjustment of the pH under constant flushing with Ar with NaOH (CO2 free) and final flushing with purified Ar on top before closing the test tubes and sealing them with parafilm. The test tubes were automatically shaken for 5 days. After centrifugation (Hettich EBA 8) for 30 min at 7,000 g, the pH was measured in the supernatant (Orion Ross semi-micro combination electrode). In some cases, the pH was also measured before centrifugation, but no significant differences were found. The pH measurement set-up was calibrated using at least five commercial buffers (Merck Titrisol, kept in refrigerator prior use). All the measurements were done on the potential scale and under constant flow of Ar over the samples. After centrifugation a volume of 3 ml of the supernatant was pipetted into polyethylene tubes (Sarstedt, 5 ml) and analysed by ICP-MS (ICP-MS Elan 6100) for U(VI) and Si. The amount of adsorbed U(VI) was calculated from the difference between measured U(VI) and known total initial U(VI). 2.4 Thermodynamic Simulations To predict the results of batch experiments, a Three-Site Non-Electrostatic surface complexation Model (TSNEM, Kohler et al. 1996) was applied. This model had been calibrated on results of column experiments (at pH 4–4.9) and a series of batch results both with a sample of MUS 30 in 0.01 M NaNO3. The smaller surface area reported by Kohler et al. (1996) is due to their effort to separate the fines since they used the sample for column experiments and had to avoid losses of small particles from the column. The development of the surface complexation model by Kohler et al. was a step-wise process. To begin with, a one-site approach was tested followed by a two-site approach, which both failed to describe the experimental results. The use of an additional site resulted in good agreement with experiment. Table 2 summarises the complete set of model parameters for the aqueous speciation and the surface complexation reactions. The set of constants has to be used in a self-consistent way, so that it should be limited to the same ionic medium used by Kohler et al. (1996). In this model, uranyl–silica complexes are not considered. We performed separate speciation calculations on this issue with a complete set of stability constants from Nuclear Energy Agency Thermodynamic Database, including associated auxiliary data (NEA 2003). It should be noted that the applied model, which was developed by Kohler et al. (1996) is purely based on best-fit parameters. In this study, no attempt was made to adjust any of these parameters. We also stress that we do not attempt to associate any mechanistic meaning to the adsorption model, neither with respect to the nature of the sites nor with respect to the postulated surface species. We refer to Goyne et al. (2002) for a discussion of the issue of site densities. Given that the full model by Kohler et al. (1996) is accepted as such, one crucial particle property is the specific surface area. As discussed above, there is some uncertainty about the measured surface areas. For the Fil-Pro sample this has been discussed elsewhere (Lu¨tzenkirchen and Huber 2007). Since great care had been taken to measure the areas with sufficient surface, we accept the measured values. Making use of these values, the respective site concentrations can be calculated for the modelling using the site densities in Table 2. The multi-component speciation code ECOSAT 4.8 (Keizer and van Riemsdijk 1998) was applied for all the modelling calculations.
123
448
Aquat Geochem (2009) 15:443–456
Table 2 Thermodynamic data for aqueous phase reactions and surface complexation reactions (Kohler et al. 1996) Species
Reaction
[UO2OH]?
H2O(l) ? UO22? $ [UO2OH]? ? H?
UO2(OH)2(aq)
2H2O(l) ? UO22? $ UO2(OH)2 ? 2H?
-12.00
[UO2(OH)3]-
3H2O(l) ? UO22? $ [UO2(OH)3]- ? 3H?
-20.00
4H2O(l) ? UO22? $ [UO2(OH)4]2- ? 4H?
-32.00
[(UO2)2(OH)]3?
H2O(l) ? 2UO22? $ [(UO2)2OH]3? ? H?
-2.80
[(UO2)2(OH)2]2?
2H2O(l) ? 2UO22? $ [(UO2)2(OH)2]2? ? 2H?
-5.63
[(UO2)3(OH)4]2?
4H2O(l) ? 3UO22? $ [(UO2)3(OH)4]2? ? 4H?
-11.90
[(UO2)3(OH)5]?
5H2O(l) ? 3UO22? $ [(UO2)3(OH)5]? ? 5H?
-15.56
-
[(UO2)3(OH)7]
7H2O(l) ? 3UO22? $ [(UO2)3(OH)7]- ? 7H?
-31.20
[(UO2)4(OH)7]?
7H2O(l) ? 4UO22? $ [(UO2)4(OH)7]? ? 7H?
-21.90
[UO2NO3]?
NO3- ? UO22? $ [UO2NO3]?
[UO2(OH)4]
2-
Log K (I = 0) -5.20
0.26 Log K (I = 0.01)
fa
?
-5.04
0.9748
S2OH ? UO22? ? H2O $ S2OUO2OH ? 2H?
-3.56
0.0250
Surface complexation reactions S1OH ? UO2
2?
? H2O $ S1OUO2OH ? 2H
S2OH ? UO22? $ S2OUO2? ? H? S3OH ? UO22? ? H2O $ S3OUO2OH ? 2H?
0.642 -1.15
0.0250 0.0002
Total site density = 0.35 lmol sites/g of quartz (1.06 lmol sites/m2) a
f = fraction of the total surface site density for each site type
3 Results and Discussion 3.1 Batch Adsorption Experiments Figure 1 displays the results of the batch experiments for the different samples as percentage adsorbed U(VI) versus pH. At low pH (\4, quartz is uncharged under this condition), adsorption is weak but increases in the range of pH 4–5 sharply (from *10% to 80–90%). From pH 5 to 10, adsorption is constant at almost 100% showing that there is no influence of CO2. The presence of CO2 would cause a decrease of adsorption with increasing pH because of the formation of dissolved uranyl–carbonato complexes. There is no significant influence of the ‘‘nominal’’ grain size detectable for the MUS samples, which is not expected due to the high amount of fines controlling the behaviour or due to identical adsorption properties of the different particles. Also for the two other pretreated samples the pH-edge is between pH 4 and 5. Thus, the results for all the pre-treated samples agree (pH edge in the same pH range and no influence of CO2). The results of the batch experiments for the untreated sea-sand differ from those described above. Adsorption starts at lower pH values. Independent results of a column experiment with the untreated sea-sand sample at pH 4 support this observation since no breakthrough of uranyl was detected during the sampling period (9 h) while column
Fig. 1 Results of the batch experiments (open circles) and corresponding calculations with ECOSAT (solid c lines) for a Min-U-Sil 5, b Min-U-Sil 10, c Min-U-Sil 15 and d Min-U-Sil 30, e Min-U-Sil 40, f Fil-Pro 12/20, g sea-sand untreated and h sea-sand pre-treated
123
Aquat Geochem (2009) 15:443–456
449
123
450
Aquat Geochem (2009) 15:443–456
experiments under the same conditions with the pre-treated FP sample showed a breakthrough. Preliminary streaming potential measurements on the two samples suggest that the treated sample is behaving ‘‘quartz’’-like, since no isoelectric point was found for pH [ 4 and a typical levelling off found for some silica samples occurred at pH \ 4 at near-zero zeta potential, whereas the treated sample had a sharp isoelectric point of about 4.8, which is typical for coated silica samples. These results would suggest that the treated sample macroscopically behaves like quartz/silica. 3.2 Dissolved Silica Concentrations Measured silica concentrations for 5-day equilibration periods are shown in Fig. 2 for all samples. In general, the lowest solubilities were obtained for acidic conditions at *0.1 mM Si rising to *1 mM Si for high pH values. As expected and in agreement with the literature (Iler 1979), finer quartz samples (MUS) showed higher aqueous Si concentrations than the coarser samples. Figure 2b also includes measured dissolved Si concentrations for MUS 10 in the presence of U(VI). They are as expected in the range of data in the absence of U(VI). The measured concentrations can be safely used to evaluate the impact of the dissolution on the speciation of U(VI) in solution. It is surprising that aqueous silica concentrations are higher than quartz equilibrium concentrations. This is discussed in more detail in the subsequent section. 3.3 Results of the Thermodynamic Simulations Figure 1 includes the results of blind predictions by the TSNEM. The model predicts the new experimental data for the MUS quartz samples quite well, which may have been expected due to the calibration of the model with MUS 30 by Kohler et al. (1996) although in a sample without fines. Predictions of the behaviour of the pre-treated coarser quartz samples were less successful. Surprisingly, the data for the untreated sea-sand sample are nicely predicted. This indicates the importance of checking the surface composition of such samples by spectroscopic methods. Without the XPS or EDX results, the untreated sea-sand would have been handled as a quartz sample, a model would have been found to be quite successful (Fig. 1g), and the underlying problems would not have been discovered. For the pre-treated SS sample, the tendency with increasing solid content is captured,
Fig. 2 a Experimentally determined solubilities for all the quartz materials used in the batch experiments (symbols; I = 0.01 M) and calculated solubilities for amorphous (solid line) and crystalline (dashed line) quartz; b Comparison of the quartz solubility for all the Min-U-Sil sorts in the presence of uranyl (crosses), and for Min-U-Sil 10 (black circles) in the absence of uranyl
123
Aquat Geochem (2009) 15:443–456
451
but the pH-range of the predicted pH-edge is too low. The reason for the failure in the predictions for the pre-treated coarse samples might be uncertainties in the surface area measurements. For the FP sample, the lower value for the surface area is used. A slight increase in the value (to that found from the full isotherm) shifts the curve to lower pH. The variation in the available values for the specific surface areas of these samples indicates that the uncertainty in this important property is high. Moreover, for the sea-sand samples, the increase in specific surface area after pre-treatment was somewhat unexpected. The low values involve some uncertainty although the amount of surface in the measurement set-up was deemed sufficient. The data can be exactly described using specific surface areas between calculated geometrical and measured values. The predictions using the measured values lie within one pH unit of the experimental results. Calculated equilibrium solubilities for amorphous and crystalline quartz are compared with measured concentrations in Fig. 2a. The experimentally determined concentrations are intermediate between the two theoretical curves. All the data are consistently higher than the equilibrium quartz solubility, but dissolution has not reached equilibrium with amorphous silica. The results suggest that the surface of quartz transformed to a surface corresponding to amorphous silica in agreement with earlier suggestions that the high charging of some quartz sample indicates a gel-like layer (Riese 1982). This experimental results suggest that it is necessary to measure silica concentrations in solutions instead of relying on equilibrium solubilities when estimating the effect of dissolved silica (here, this is important for uranyl–silicato complexes). Figure 3 shows the results of speciation calculations for the aqueous uranyl–silica system for equilibrium with respect to amorphous silica and quartz. Experimental results are between quartz and amorphous silica. The fact that aqueous silica is above the quartz equilibrium concentrations may have different reasons. One would be that pre-treatment like grinding causes an amorphisation of the surface, which then results in interfacial behaviour of amorphous silica. Another possibility to explain this effect would be the often-suggested formation of the gel-like layer when quartz is in contact with water. It should be noted that there is no general agreement in the scientific literature about the existence of such a gel layer or an amorphous surface (see Ball et al. 1997 and references therein). The calculations indicate that a refined model should consider the formation of uranyl–silicato complexes, which appear or dominate in the mildly acidic pH-range.
Fig. 3 Calculated speciation of uranyl in the presence of dissolved silica for a equilibrium with quartz, b equilibrium with amorphous silica. Only major species are shown. All the thermodynamic data are adopted from the most recent NEA compilation (Vol. 5)
123
452
Aquat Geochem (2009) 15:443–456
3.4 Comparison of the Experimental Results with Literature Data Figure 4 shows a comparison of our adsorption data for the seven pre-treated samples with literature data for quartz and amorphous silica. Data are shown as distribution coefficients calculated with the surface area of the respective sorbent to include the effects of specific surface area. It can be expressed as follows: CU ðTOTU ½UÞ TOTU =/m ¼ 1 =/m; ¼ Kd ¼ ½U ½U ½U where Kd = Distribution coefficient in l/m2, CU = Surface excess of U(VI) in mol/m2, [U] = Dissolved U(VI) concentration in mol/l, TOTU = Total U(VI) concentration in mol/l, / = Surface area in m2/g, m = Solid concentration in g/l. It is important to note that the uncertainty in the distribution coefficient close to 100% adsorption is typically getting high. Moreover, at low pH where most of the uranyl is remaining in solution, the relative error (difference between two big numbers) becomes important. Literature data in Fig. 4 may be in the presence or absence of carbonate. For the low pH range, carbonate complexation can be neglected. Since carbonate does not adsorb to silica in significant amounts as was shown experimentally by Vuceta (1976), the surface charge of pre-treated samples (i.e. largely or entirely Fe-free based on XPS results) is not affected by carbonate. The results of our study agree fairly well with published results. Deviations for data at high pH values ([8) are due to the presence of carbonate in some literature data
Fig. 4 Comparison of literature data for uranyl adsorption with silica polymorphs. Data are represented as a surface specific distribution coefficient (Kd/[l/m2]). Only data for ionic strength of 0.01 M are used
123
Aquat Geochem (2009) 15:443–456
453
resulting in a decrease of log Kd due to uranyl-carbonato complexation. Interestingly, in this plot the SS (treated) and FP samples fit well. Instead, the data for untreated SS sample are somewhat lower than the bulk of the data. This would suggest that the success of the model for the untreated SS sample is accidental and that the model should be considered with care when transferring it from one solid to another. Note that the literature Kd values span about one pH unit. So if the experimental data were reliable, then a deviation of one pH unit can be tolerated. However, such a deviation can, in particular within a model prediction, result in significant over- or underprediction of uranyl retention. From this point of view such a deviation is undesirable. 3.5 Surface Speciation Figure 5 shows the surface speciation predicted by the applied surface complexation model (TSNEM). The figures also include the amount of uranyl relative to the surface area in the system. The inherent site densities of the three surface sites of the model are a fraction of 1.06 lmol/m2, namely 97.48%, 2.5% and 0.02% for sites 1, 2 and 3, respectively. For none of the conditions, the high-affinity site is present in sufficient amount to fix all of the uranyl present. For the experiments with the high uranyl-to-surface area ratios, a common sequence in the surface speciation occurs: at low pH values (\4) where the overall adsorption is low (max. *20%) and the free uranyl ion is the major aqueous species, S2OUO2? is the dominating surface species. This is somewhat unexpected since S3 is the high-affinity site, but since only a hydrolysed uranyl surface species was postulated on that site by Kohler et al. (1996), the predominant adsorption on S2 is caused by the pHdependence. S2OUO2? rarely reaches up to more than about 10%, and is negligible at pH [ 5. Above pH 4, S1OUO2OH becomes the major surface species over the whole remaining pH range ([4–10) at about 40% except for pre-treated SS at the lowest uranylto-surface ratio where S3OUO2OH dominates (Fig. 5h). S3OUO2OH reaches values up to *20% except for the uranyl-to-surface area ratios below 0.3 nmol m-2 can be understood from comparing this ratio to the site density. S2OUO2OH usually plays an intermediate role (contributing about *30%) and is almost dominated by S1OUO2OH. This can be understood by considering the product of site density and adsorption constant, which is higher for S1 than for S2 for the hydrolysed surface species by a factor of about 1.3, which reflects fairly well the calculated differences. In summary, saturation of the strong sites occurs even for the low uranyl concentrations and the sites with ‘‘weaker’’ affinity dominate, whereas at higher pH site 1 dominates site 2 although site 2 is not saturated. The role of site 2 as a ‘‘more’’ reactive site is only in the postulated S2OUO2? species, which always dominates at low pH. For most of our batch experiment conditions, the model predicts similar contributions from all the four surface species at the onset of adsorption (around pH 4) where most column experiments were done, but for the latter the solid-to-liquid ratio is even higher, so that it is expected that the high-affinity site will be more important.
4 Summary and Conclusions The adsorption of uranyl onto seven different samples of quartz was studied by static batch experiments. Thermodynamic predictions were carried out to simulate the experimental results using the complete surface complexation model calibrated on adsorption of U(VI) onto Min-U-Sil 30 as published by Kohler et al. (1996).
123
454
123
Aquat Geochem (2009) 15:443–456
Aquat Geochem (2009) 15:443–456
455
b Fig. 5 Surface species distribution of the surface complexation model (TSNEM) for the corresponding batch experiments. a MUS 5, b MUS 10, c MUS 15, d MUS 30, e MUS 40, f Fil-Pro 12/20, g sea-sand (treated) at 0.18 kg/l, and h sea-sand (treated) at 0.6 kg/l
The applied SCM describes well the results of the batch experiments for the MUS samples given that the experimental data were obtained in a different laboratory and that the samples used in this study included significant amounts of fine particles. Comparison of our results with literature data gives confidence in the measurements under various experimental conditions (in terms of procedures such as solid liquid separation or uranyl analysis). In contrast to the model predictions, the measured Kd values for the pre-treated samples are consistent with literature data and with each other while the data for the untreated sample deviates. Prediction of the data for the coarser samples is satisfactory but off by a pH unit. This may be due to uncertainty in the measured specific surface area which is higher than expected based on grain size. For the untreated sea-sand sample, a perfect prediction was obtained although this should not be the case due to the impurities on this sample. Experimentally determined silica concentrations for the seven samples are in the range from 0.1 mM to about 1 mM. No solubility equilibrium was attained within 5 days as inferred from comparison of the measured values with calculated equilibrium solubilities for amorphous silica and quartz. Measured values of dissolved silica concentrations are higher than equilibrium quartz solubilities. This implies some amorphisation of the surfaces either due to grinding of the raw material or due to formation of gel-like layers in contact with water. The solubility measurements along with speciation calculations for uranyl indicate that uranyl–silicato complexes need to be considered and that silica dissolution needs to be measured if ambiguity with respect to the total dissolved silica in the aqueous speciation scheme is to be avoided. Surface speciation calculations within the model used show that depending on the uranyl-to-surface ratio, the high-affinity site is often saturated in batch experiments. The situation is expected to be different in column experiments, where that ratio is very low. The use of the unhydrolysed surface species with an intermediate site density causes adsorption on that site only at the low pH range. Acknowledgements We thank A. Kaufmann and F. Geyer for ICP-MS analysis, D. Schild for XPS measurements, T. Kisely for BET analysis (all of them from INE) and T. Wiss and H. Thiele (both JRCITU) for EDX and SEM investigations. A thorough reading of the manuscript by Dr. B. Kienzler (INE) is gratefully acknowledged.
References Artinger R, Rabung T, Kim JI, Sachs S, Schmeide K, Heise KH, Bernhard G, Nitsche H (2002) Humic colloid-borne migration of uranium in sand columns. J Contam Hydrol 58:1–12. doi:10.1016/S01697722(02)00032-3 Ball MC, Brown DS, Perkins LJ (1997) Diffuse layers on the surface of mineral quartz. J Mater Chem 7:365–368. doi:10.1039/a603570a Davis JA (ed) (2001) Surface complexation modeling of uranium(VI) adsorption on natural mineral assemblages, Report NUREG/CR-6708, US Nuclear Regulatory Commission W D.C. Gabriel U, Charlet L, Schla¨pfer CW (2001) In situ speciation of uranium (VI) at the silica–water interface: a combined TRLFS and surface complexation study. In: Hellmann R, Wood SA (eds) Water–rock interactions, ore deposits, and environmental geochemistry: a tribute to David A. Crerar. Geochemical Society, St. Louis, Spec. Publ. No. 7
123
456
Aquat Geochem (2009) 15:443–456
Goyne KW, Zimmerman AR, Newalkar BL, Komarneni S, Brantley SL, Chorover J (2002) Surface charge of variable porosity Al2O3(s) and SiO2(s) adsorbents. J Porous Mater 9:243–256. doi:10.1023/ A:1021631827398 Grenthe I, Fuger J, Konings RJM, Lemire RJ, Muller AB, Nguyen-Trung C, Wanner H (1992) Chemical thermodynamics of uranium. Elsevier Science, Amsterdam Hongxia Z, Zuyi T (2002) Sorption of uranyl ions on silica: effects of contact time, pH, ionic strength, concentration and phosphate. J Radioanal Nucl Chem 254:103–107. doi:10.1023/A:1020801816866 Iler K (1979) The chemistry of silica: solubility, polymerization, colloid and surface properties and biochemistry. Wiley, New York Jung J, Hyun SP, Lee JK, Cho YH, Hahn PS (1999) Adsorption of UO22? on natural composite materials. J Radioanal Nucl Chem 242:405–412. doi:10.1007/BF02345570 Keizer MG, van Riemsdijk WH (1998) ECOSAT: technical report of the department of soil science and plant nutrition, Wageningen agricultural. University of Wageningen, The Netherlands Kohler M, Curtis GP, Kent DB, Davis JA (1996) Experimental investigations and modelling of uranium(VI) transport under variable chemical conditions. Water Resour Res 32:3539–3551 Lu¨tzenkirchen L, Huber F (2007) Heterogeneities in adsorption of aqueous solutions—an example of the effect of surface coverage. Adsorpt Sci Technol 25:503–516 Mibus J, Sachs S, Pfingsten W, Nebelung C, Bernhard G (2007) Migration of uranium (IV)/(VI) in the presence of humic acids in quartz sand: a laboratory column study. J Contam Hydrol 89:199–217. doi: 10.1016/j.jconhyd.2006.08.005 NEA (2003) Vol. 5. Update on the chemical thermodynamics of uranium, neptunium, plutonium, americium and technetium. OECD Nuclear Energy Agency, Paris Pathak NP, Choppin GR (2006) Kinetics and thermodynamics of uranium(VI) sorption on hydrous silica. Radiochim Acta 95:507–512. doi:10.1524/ract.2007.95.9.507 Pathak NP, Choppin GR (2007a) Sorption of uranyl on hydrous silica: effects of ionic strength and ethylenediaminetetraacetic acid (EDTA). J Radioanal Nucl Chem 272:37–43. doi:10.1007/s10967006-6783-6 Pathak NP, Choppin GR (2007b) Sorption of uranyl(VI) cations on suspended silicate: effects of N-donor ligands, carboxylic acids, organic solvents, and metal ions. Radiochim Acta 95:245–250. doi:10.1524/ ract.2007.95.5.245 Pokrovsky OS, Golubev SV, Mielczarski JA (2006) Kinetic evidences of the existence of positively charged species at the quartz-aqueous solution interface. J Colloid Interface Sci 296:189–194. doi:10.1016/ j.jcis.2005.09.001 Riese AC (1982) Adsorption of radium and thorium onto quartz and kaolinite: A comparison of solution/ surface equilibria models. Ph. D. dissertation, Colorado School of Mines Vuceta J (1976) Adsorption of Pb(II) and Cu(II) on a-quartz from aqueous solutions: influence of pH, ionic strength and complexing ligands. Ph. D. dissertation, Calif. Inst. of Technol., Pasadena Zuyi T, Taiwei C, Jinzhou D, XiongXin D, Yingjie G (2000) Effect of fulvic acids on sorption of U(VI), Zn, Yb, I and Se(IV) onto oxides of aluminium, iron and silicon. Appl Geochem 15:133–139. doi: 10.1016/S0883-2927(99)00042-6 Zuyi T, Taiwei C, Weijuan L, Jinzhou D, Xiongxin D, Yingjie G (2004) Cation adsorption of NpO2?, UO22?, Zn2?, Sr2?, Yb3?, and Am3? onto Oxides of Al, Si and Fe from aqueous solution: ionic strength effects. Colloid Surf A 242:39–45. doi:10.1016/j.colsurfa.2004.04.044
123
