Introduction

Uranium is of increasing environmental concern due to a grown awareness of its risk to soil and water. Elevated concentration of uranium can be related to uranium milling and mining sites (Morrison and Cahn 1991), nuclear fuel and nuclear weapons production sites (Riley et al. 1992), combustion of coal and oil, in particular, when no proper combustions gas cleaning systems are installed, and the application of phosphate fertilizers (Barisic et al. 1992; Zielinski et al. 2006). A provisional drinking water MCL for uranium of 15 μg L−1 has been established by the World Health Organization (WHO 2004). However, it is discussed worldwide that 10 or 5 μg L−1 would be more reasonable (BFR 2005; Hickox and Denton 2001; Kurttio et al. 2002; Raymond-Whish et al. 2007). Geochemical processes occurring naturally, including dissolution/precipitation, redox reactions, and sorption/desorption reactions at the water–rock interface, control the mobility and transport of uranium in the subsurface system, such as aquifer sediments, soils, and groundwater.

Sorption is an important removal mechanism that controls uranium concentration in groundwater (Prikryl et al. 2001). However, the dependence of sorption on aqueous solution properties (e.g., pH, Eh, total uranium concentration, ionic strength, and presence of complexing ligands) and characteristics of sorbing materials (e.g., mineral compositions, surface area, density of sorptive surface sites, solid/solution ratio) makes the prediction of uranium retardation difficult (Prikryl et al. 2001). The sorption of uranium in natural water, on geological material typically increases with increasing pH up to the neutral region, while in system equilibrated with air, a significant decrease of sorption above pH 7 due to a stable uranyl-carbonate complexation is observed (Payne 1999).

Among common sorbing materials are clay minerals, which have a very high retention capacity due to (1) their large surface area as well as their osmotic swelling capacity and consequently their plasticity and impermeability and (2) the property of these minerals to simultaneously create a permanent negative charge within the structure and a variable charge at the particle edge. The negative surface charge is created due to the extensive isomorphous substitution of cations in the lattice by cations of lower charge (e.g., the substitution of Al3+ for Si4+ in the tetrahedral sheet and Mg2+ for Al3+ in the octahedral sheet). Positive charge may occur due to the edge defects (e.g., protonation of broken Al–OH bonds exposed at particle edges). On the other hand, kaolin minerals do not have extensive isomorphous substitution of metals in tetrahedral and octahedral sheets. Thus, edge defects are more important sources of positive charge than isomorphic substitution (Lin and Puls 2000). Hence, clay minerals can sorb cations and/or anions that neutralize the structure charge. Some of the most important clays are montmorillonite and kaolinite, both because of their prevalence in soil and ground water environments and because they have been recognized as fundamental for retarding the migration of many contaminates in soils, sediments, and rocks (Bachmaf et al. 2008; Borovec 1981; Catalano and Brown 2005; Hyun et al. 2001; Olguin et al. 1997; Payne et al. 2004). They are as well considered as a barrier material for radionuclide retention in nuclear waste repositories (Lajudie et al. 1995; Neall et al. 1995).

Sorption of uranium by clay minerals has been extensively documented in the literature. However, most of the previous sorption investigations have been performed at relatively high uranium concentration (>100 μg L−1). Using the geochemical code PHREEQC (Parkhurst and Appelo 1999) with the updated Nuclear Energy Agency thermodynamic database (NEA 2007) (Grenthe et al. 2007), positive saturation indices were found for certain uranium minerals (e.g. Schoepite) when the total uranium concentration was higher than 100 μg L−1, in particular, at pH above 5. Furthermore, X-ray characterizations for kaolinite treated with higher concentration of uranium showed evidence of uranium mineral precipitation on the surface of clays. Thus, it is likely that amorphous phases have been precipitated in studies using uranium concentrations above 100 μg L−1. The practical relevance for conducting experiments at uranium concentration below 100 μg L−1is that uranium occurs naturally in low concentrations (a few parts per million) in soil, rock as well as in surface and groundwaters (Bernhard 2005; Langmuir 1997).

The purpose of this investigation was to examine the sorption of uranium with five different types of clay minerals as a function of pH and varying NaCl concentrations. Monodentate and bidentate uranyl binding (as UO2 2+ cation and various monomeric or polymeric hydrolyzed species) to permanent charged surface sites as well as to edge hydroxyl groups of clay minerals were considered to be important in sorption modeling approaches (Pabalan et al. 1996; Prikryl et al. 1994). Specific surface area, clay composition, and surface site density were evaluated in these models, and the corresponding equilibrium constants for various modes of binding were calculated.

Different models have been employed to elucidate uranyl sorption onto clay minerals. Ion exchange modeling used by (Fletcher and Sposito 1989) was an early approach. A more recent and sophisticated approach is surface complexation modeling (Kowal-Fouchard et al. 2004; McKinley et al. 1995; Pabalan et al. 1996; Turner et al. 1996; Zachara and Mckinley 1993). Although previous researchers assumed that the uranyl sorption by clay minerals at acidic to neutral pH values is dominated by ion exchange reactions, recent works e.g., (Arda et al. 2006; Chisholm-Brause et al. 2004) have shown that the reactivity of metal-oxides like edge sites in clay minerals is responsible for uranium sorption over a wide range of pH and electrolyte concentration. A modified multisite surface complexation model developed by (Zachara and Mckinley 1993) for montmorillonite, and that developed by (Payne et al. 2004) for kaolinite was used to explain the mechanisms controlling the uranium sorption.

Materials and methodology

Clay minerals characterization

Selected reference clay minerals, well-crystallized kaolinite (KGa-1b), poorly crystallized kaolinite (KGa-2), Ca-montmorillonite (STx-1b), and Na-montmorillonite (SWy-2) were supplied from the Clay Minerals Society (CMS) Source Clays Repository (University of Purdue, West Lafayette). Morocco natural bentonite (IBECO) was obtained from the Federal Institute for Geosciences and Natural Resources (BGR) in Hannover (Germany). Powder X-ray diffraction (XRD) analyses were used to characterize the mineralogical composition of the phyllosilicate clay minerals according to (Chipera and Bish 2001; Ufer et al. 2004). The results show that KGa-1b and KGa-2 contain predominantly titanium as a trace of anatase (Pruett and Webb 1993). Standard STx-1b and SWy-2 montmorillonite and natural bentonite are mainly composed of smectites (67, 75, and 80%, respectively) and small amounts of impurity phases, such as quartz, feldspar, kaolinite, plagioclase, orthoclase, and opal. The chemical compositions of the five materials are summarized in Table 1, which shows the presence of approximately 1.7% TiO2 in each of the Georgia kaolinite samples. Specific surface areas and cation exchange capacities of the clay sorbents are provided from the previous studies and given in Table 1. Those authors determined the specific surface area by multi-point Brunauer- Emmett-Teller (BET) N2 sorption with a Quantisorb Jr. surface area analyzer (Quantachrome Corporation, Syosset, NY). The procedures performed to measure the cation exchange capacity are explained by (Borden and Giese 2001). The intention of this research was to use the source clay materials without any pretreatment to preserve their natural mineralogy.

Table 1 Properties of clay minerals (chemical composition in %, cation exchange capacity, and specific surface area)
Full size table

Uranium solutions and analytical procedures

Uranium stock solutions of 100 μg L−1 (4.2 × 10−7 M) were prepared by dissolving uranyl nitrate hexahydrated UO2 (NO3)2.6H2O (Chemapol, Germany) in Milli-Q ultrapure water (18 MΩ/cm). All experiments were carried out in a matrix of sodium chloride (Merck, Darmstadt-Germany) at constant values of 0.001, 0.01, and 0.025 M, respectively. The pH of each test solution was adjusted to the required value with diluted NaOH or HCl solutions (Merck, Germany) using a combined glass electrode (WTW GmbH, Germany). The determination of total uranium was carried out by differential pulse adsorptive cathodic stripping voltammetry using a hanging mercury drop electrode (HMDE) as working electrode (797 Va Computrace, Methrohm, Switzerland). The technique is based upon adsorptive accumulation of the metal ion complexed with the ligand chloranilic acid at the electrode surface and then scanning the potential on the working electrode in the negative direction. The detection limit of the voltammetry was determined to be 0.05 μg L−1. All experiments and all analytical determinations were performed in duplicates. The maximum experimental error was ±3%.

Experimental procedure

All sorption batch experiments were carried out by reacting 1.5 g of the respective clay mineral with 900 mL of the 100 μg L−1 uranium(VI) solution with 0.001, 0.010, and 0.025 M NaCl in 1 L PTFE vessels under ambient atmosphere (Pco2 10−3.5 hPa). The pH was monitored in all experiments and adjust when necessary with 0.1 M NaOH or 0.1 M HCl (Merck, Germany) using an automated titrator (794 Basic Titrino, program version 5.794.0010, Methrohm, Switzerland) in “pH-stat” mode. For pre-equilibration, the suspension was continuously stirred using a magnetic stirrer (IAK®, Germany). For kinetic experiments, samples were taken and centrifuged after 0.5, 1, 3, 6, 12, and 24 h. For the majority of the batch experiments, samples were taken after 24 h and the solution was then separated from the solid by centrifugation (20 min, 8,000 rpm, MLW Medizintechnik). The supernatants of all individual samples were filtered (0.2 μm cellulose acetate filters, Membrex, Germany) and immediately analyzed by voltammetry for uranium and by ion chromatography (850 Professional IC, Methrohm, Switzerland) for Na+, Ca2+, and Mg2+. Total Fe and Si4+ were determined by photometry (Hach, USA). For XRD characterizations, the residual solids were dried in an oven at temperature below 45°C for 1 month and crushed with a mortar and pestle to pass a 40 μm sieve. The percentage of sorbed uranium was calculated from the difference between the initial and final uranium concentration using the following Eq. (1):

$$ {\text{Sorption (\%) = }}\left( {{\frac{{C_{\text{i}} - C_{\text{f}} }}{{C_{\text{i}} }}}} \right) \times 100, $$
(1)

where C i and C f are the uranium concentrations in the initial and final solutions, respectively.

Blank experiments carried out without a mineral sorbing phase indicated substantial uptake of uranium on glass container walls, particularly in the pH range from 4 to 9 and at low uranium concentration. Therefore, all tools used in the experiments (e.g., vessels, centrifugation tubes, magnetic stirrer) were made of PTFE, which exhibited no reaction with uranium solution (<1% uptake on container walls for all pH). Supplementary experiments were conducted to estimate the influence of filtration on results. The supernatants were analyzed for uranium prior and after filtration and it was shown that cellulose acetate filters have no effect on the results.

Results and discussion

Uranium aqueous speciation

Uranium aqueous speciation were computed for the solution with a total uranium of 100 μg L−1 in a background electrolyte of 0.01 M NaCl equilibrated with atmosphere (i.e., partial pressure of 10−3.5 hPa for Pco2). The distribution of aqueous uranyl species was calculated with the geochemical code PHREEQC-2 (Parkhurst and Appelo 1999) with Nuclear Energy Agency (NEA 2007) thermodynamic database. Uranium speciation are presented in Fig. 1. Free uranyl ion UO2 2+ is the dominant species in the acidic pH range up to 5. With increasing pH, uranyl ion becomes more hydrolyzed and forms hydrolyzed uranyl species such as UO2OH+. In the equilibrium with atmospheric levels of CO2, the most predominant aqueous uranium species under neutral and alkaline conditions are uranyl-carbonate-complexes ((UO2)2CO3(OH) −3 , UO2(CO3) 2−2 , and UO2(CO3) 4−3 ).

Fig. 1
figure 1

Uranium speciation in 0.01 M NaCl, ΣU = 100 μg L−1, Pco2 = 10−3.5 hPa, 25°C, calculated with the geochemical code PHREEQC using NEA_2007 database

Full size image

Kinetic rates of sorption

The aim of kinetic experiments was to examine the time required for uranium sorption equilibration. The results for sorption of uranium on clay minerals at constant pH of 4, 6, and 8 are shown in Fig. 2. The results for other pH were similar and are not shown here. Only the results for KGa-2 are presented, because the other kaolinite showed a very similar behavior. The general observation in Fig. 2 indicated that under neutral conditions both smectites and kaolinites exhibited a classic biphasic sorption of uranium which is characterized by an initial rapid uptake over the first 30 min followed by a slower increase in the uranium sorption over the reminder of the experiments. In contrast, under acidic and alkaline conditions, smectites demonstrated an initially fast uranium uptake followed by a slower decrease in the uranium sorption. The much faster sorption for SWy-2 than other smectites is due to the high content of iron, which can rapidly interact with uranium forming inner-sphere complexes. The second slower step of sorption is due to interparticle and intraparticles diffusions in pores. At pH 6 and 8, the sorption curve of KGa-2 reaches a maximum after 30 min. Then, it appears stable over 24 h, whereas 5 h were required to reach the sorption equilibrium at pH 4. This tendency suggests the gradual saturation of the surface sites at increasing experimental time: at neutral conditions, uranium is more effectively bound and thus correspondingly retained more rapidly. Comparison between kinetic sorption data for all smectites observed at pH 8 reveals that sharp decrease in uranium sorption occurred after 30 min from the beginning of the sorption experiments. This can be explained by the formation of uranyl-carbonate complexes, which are weakly binding to the surface of smectites. Yet, this phenomenon was not seen for uranium sorption on kaolinites, which can be attributed to the presence of anatase as impurity phase. The results clearly show that data taken from routinely done titration experiments with only 10 or 15 min equilibrium time might be biased. However, even 24 h was observed as under estimating time for equilibrium in all cases as well.

Fig. 2
figure 2

Kinetic sorption of uranium onto kaolinite (KGa-2), montmorillonite (SWy-2), montmorillonite (STx-1b), and natural bentonite (IBECO). Sorption experimental conditions: ΣU0 = 100 μg L−1, I = 0.01 M, suspension density = 1.66 g L−1, Pco2 = 10−3.5 bar, 25°C, reaction time = 24 h

Full size image

Effect of pH and Na+ concentration on U(VI) sorption

The effect of pH and the Na+ concentration on the uranium sorption behavior onto the selected clay minerals is illustrated in Fig. 3. General similarities between the two kaolinites (KGa-1b and KGa-2) include lower uranium sorption at pH 4 and much greater uranium uptake than other three clay minerals in the pH range 5–9. Minor ionic strength dependency was exhibited for kaolinites at pH above 6. Conversely, the sorption of uranyl by smectites (IBECO bentonite, Na-montmorillonite, and Ca-montmorillonite) varied with pH and ionic strength. Sorption was most dependent on the ionic strength at low pH. The high sorption capability of SWy-2 observed at pH 4 is resulted from the high Na+ content in its structure. This can suppress the competition between uranyl ions and Na+ electrolyte and allow uranium species to bind to the surface sites of SWy-2. By comparison, competitive effect of sodium with uranium species can minimize the sorption of uranium onto STx-1b and IBECO bentonite at pH 4. This assumption is supported by (Suter et al. 2008), who demonstrated that sodium can form an inner sphere complex on the surface of smectites. However, for the pH range from 5 to 9, differences between uranium sorption on all clay minerals due to the variations in ionic strength were typically less than 15%. These observations did agree well with those reported by (Hayes et al. 1988) and (McBride 1997). They found that ions that form outer-sphere surface complexes exhibit reducing sorption with increasing solution ionic strength, while ions that form inner-sphere surface complexes demonstrate minor ionic strength dependence or show increasing sorption with increasing solution ionic strength. Relating to the ionic strength dependency, there are general similarities between our results and those reported by (Manning and Goldberg 1997) who found an increase of As(III) sorption on KGa-1b and SWy-1 with decreasing sodium chloride concentration at low pH, and increasing arsenic sorption with increasing sodium concentration at high pH.

Fig. 3
figure 3

Comparison of uraniumsorption data on kaolinite (KGa-1b), kaolinite (KGa-2), montmorillonite (STx-1b), montmorillonite (SWy-2), and natural bentonite (IBECO) and the surface complexation model (SCM). Experimental conditions: ΣU = 100 μg L−1, suspension density = 1.66 g L−1, Pco2 = 10−3.5 hPa, 25°C, I = 0.001, 0.01, and 0.025 M, reaction time = 24 h. Solid lines, dash lines and dot lines represent modeled data for 1, 10 and 25 mL NaCl, respectively. Open square, open circles and open triangle represent measured data for 1, 10, and 25 mL NaCl, respectively

Full size image

Indeed, kaolinic minerals such as KGa-1b and KGa-2 have Si/Al ratios of 1, whereas smectites, such as IBECO bentonite, SWy-2 and STx-1b have Si/Al ratios of 2. Based on the greater activity of aluminol sites toward uranium than silanol sites (Borovec 1981; Kohler et al. 1992), which are major surface components of phyllosilicates, it would be expected that kaolinites, that contain exposed sheets of Al octahedral, would result in greater uranium uptake by kaolinic minerals than smectites.

With pH between 6.0 and 9.0, aqueous uranium species are present as uranyl carbonate, and uranyl carbonate hydroxide (Fig. 1). In addition, maximum sorption of carbonate was attained at pH 5.5 (Zachara et al. 1987). Hence, the distinct reduced uranium sorption on smectites at higher pH is related either to the low affinity of uranium carbonate complexes to the surface sites of clay minerals or to competition between carbonate ions and uranium species. On contrary, no decrease in uranium sorption on kaolinite (KGa-1b and KGa-2) was observed when uranyl-carbonate complexes are predominant in the solution. This behavior could suggest that uranyl-carbonate species have high affinity toward anatase present as impurity phase in kaolinites.

Surface complexation modeling

A multisite layer surface complexation model was developed by Zachara and Mckinley (1993) to describe uranium sorption data by smectites. Their model is composed of pH independent exchange (X−) sites and hydroxylated edge sites for SiOH and AlOH. They assumed that at low pH and low Na+ concentration, adsorption by fixed-charge site is predominant. At high pH and high Na+ concentration, the binding to the AlOH and SiOH edge sites caused the sorption. On contrary to Zachara and Mckinley, we suppose that sorption of uranium on the selected smectites preferentially occur to the edge sites of aluminol and silanol, even under acidic conditions. Our assumption is supported by the recent spectroscopy study of (Chisholm-Brause et al. 2004), who confirmed that reactivity of metal-oxide like edge sites in clay minerals are highly responsible for uranium sorption over a wide range of pH and electrolyte concentration. This assumption is also based on the relatively high Na+ electrolyte concentration (0.01 and 0.025 M) in comparison to very low uranium concentration in the solution. This can prevent the exchange between uranyl ions and cations present in the basal plane of smectites. Furthermore, our suggestions are supported by the results of X-ray characterizations, which exhibited indistinguishable patterns for the clay minerals before and after uranium treatment. In addition, dissolution experiments performed clearly showed that similar amounts of Na+, Ca2+, and Mg2+ cations were released from each clay mineral at all pH values (data not shown). A comparison between pH for samples collected at the beginning and at the end of sorption experiments showed a slight variation (<0.5 unit) indicating only insignificant cation exchange reactions.

Earlier modeling attempts for uranium sorption onto kaolinite have involved various assumptions. The attempt of (Kohler et al. 1992) postulated that actinide binding on kaolinite only occupies the edge surface sites, which were modeled as aluminol sites. Their hypothesis was based on the results of parallel experiments with gibbsite containing only AlOH sites and quartz (only SiOH sites). This experiment yielded that sorption on quartz was much weaker. The work of (Borovec 1981) also demonstrated very little affinity of U for Si-oxide surfaces. The supposition of (Turner and Sassman 1996) exhibited that reactive kaolinite surface comprised stoichiometric proportion of silanol (SiOH) and aluminol (AlOH).

In our model, uranium sorption onto Georgia kaolinites is based on the sorption on both titanol (TiOH) and aluminol sites (AlOH). The approach was adapted from (Payne et al. 2004). Their assumptions were established upon the transmission electron microscopy equipped with energy dispersive X-ray spectroscopy (TEM/EDS) analyses and the main finding that uranium was preferentially bound to anatase present as impurity phase in kaolinite.

The equilibrium constants for acid–base surface reactions of aluminol (Eqs. 1 and 2 in Table 2) are taken from (Zachara and Mckinley, 1993), while the constants for acid–base surface reactions of silanol (Eqs. 6 and 7 in Table 2) were derived from (Mckinley et al. 1993). The non-linear least squares parameter estimation program PEST 11.8 (Doherty 2006) in conjunction with PHREEQC (Parkhurst and Appelo 1999) was used to determine the equilibrium constants for the other surface reactions given in Table 2. PEST has the advantage over FITEQL (Westall 1985), that activity coefficients can be calculated according to either extended Wateq Debye-Hückel or a SIT (specific ion interaction theory) and provides users with statistical information on the uncertainty of the estimate. Sodium sorption was also incorporated into the model (Eqs. 4, 11, 16, and 22 in Table 2) since sodium was proven to sorb onto clay minerals (Suter et al. 2008) and was used by several authors (Turner et al. 1996; Zachara and Mckinley 1993) to model uranium sorption under a broad range of ionic strength. Results of the batch experiments and the multisite surface complexation model are plotted in Fig. 3.

Table 2 Surface reactions and parameters used in the model calculations. If not indicated by Zachara and Mckinley (1993) or Mckinley et al. (1993) data were calculated with PEST and PHREEQC
Full size table

The agreement between the modeled-calculated and experimentally measured values was quantified by the root mean square error (RMSE)

$$ {\text{RMSE}} = \left[ {{\frac{1}{{n_{\text{d}} - n_{\text{p}} }}}\sum\limits_{i = 1}^{{n_{\text{d}} }} {\left( {\% \text{Re} m - \% \text{Re} \hat{m}} \right)^{2} } } \right]^{1/2} $$

where n d is the number of data points, n p is the number of adjustable parameters, i is an index, \( \% \text{Re} m \) is the percentage of uranium measured removal, \( \% \text{Re} \hat{m} \) is the percentage of uranium predicted removal. The RMSE is an estimate of the standard deviation between the predicted and the measured values expressed in dimensionless form as a fraction on the initial

The modeled data presented in Fig. 3 indicate that the model predicts the major trends of sorption of uranium with respect to pH and NaCl concentrations for the five clay minerals, and in particular for Georgia kaolinite and STx-1b montmorillonite. On the contrary an over predication of sorption can be stated on SWy-2 and IBECO for a pH range 5–7. The overestimation is probably due to the presence of impurities phases such as quartz, feldspar, plagioclase, and orthoclase, which have been not taken into account in our approach so far. The root mean square error (RMSE) obtained was 2.8, 2.3, 3.4, 12.1, and 12.1 for KGa-1b, KGa-2, STx-1b, SWy-2, and IBECO, respectively.

All fitted constants are automatically corrected to zero ionic strength with respect to either ion dissociation theory (e.g. Davies or Wateq-Debye-Hückel equation) or SIT.

Conclusions

Experimental and modeled data derived in this study indicate that sorption experiments must be performed at low uranium concentration to avoid precipitation of amorphous uranium-hydroxides. Furthermore, PTFE containers are found to be appropriate materials comparing with glass ware which appeared significant uranium sorption at higher pH.

Kinetic experiments clearly showed that the common 10 or 15 min are in many cases by far not sufficient to reach sorption equilibrium, therefore, a period of 24 h was selected as contact time to conduct the sorption experiments.

Although smectite has higher cation exchange capacity than kaolinite, both kaolinite minerals showed much greater uranium sorption than smectites in the pH range (5–9). This can be considering as an obvious evidence that surface complexation was the dominant mechanism for uranium sorption under our experimental conditions. The high uranium sorption capability for kaolinite is because it contains more exposed aluminol surface site, which have greater activity toward uranium than silanol sites.

Uptake of uranium on montmorillonites demonstrated a distinct dependency on sodium concentrations because of the effective competition between uranyl and sodium ions, whereas less significant differences in sorption were found for kaolinite. The presence of anatase as impurity in kaolinite enhanced the binding of uranyl-carbonate complexes with surface sites.

A sufficient matching between the experimental results and the modeled data was obtained using a multisite surface complexation model without assuming exchange. This model is based on the binding of the most dominant uranium species to aluminol and silanol edge sites of montmorillonite, respectively, to aluminol and titanol surface sites of kaolinite. The parameter estimation code PEST together with PHREEQC is an extremely powerful tool to estimate the equilibrium constants estimated in this work. The good agreement between measured and predicted data implies that conceptual models based on SCM approach, such as one the developed here, could be useful in extrapolating radionuclide sorption behavior over a range of geochemical conditions based on the model parameters derived from a limited set of laboratory data.