Article pubs.acs.org/est
Experimental Measurements and Surface Complexation Modeling of U(VI) Adsorption onto Multilayered Graphene Oxide: The Importance of Adsorbate−Adsorbent Ratios Thomas A. Duster,*,†,‡ Jennifer E. S. Szymanowski,† and Jeremy B. Fein† †
Department of Civil and Environmental Engineering and Earth Sciences, University of Notre Dame, 156 Fitzpatrick Hall, Notre Dame, Indiana 46556, United States ‡ Applied Chemicals and Materials Division, Material Measurement Laboratory, National Institute of Standards and Technology, 325 Broadway, Boulder, Colorado 80305, United States S Supporting Information *
ABSTRACT: Surface complexation models use experimental adsorption measurements to calculate stability constants that quantify the thermodynamic stability of adsorbed species. However, these constants are often poorly constrained due to nearly complete removal of the solute from solution and/or because the tested adsorbate:adsorbent ratios are not varied sufficiently. Using data sets that quantify the adsorption of U(VI) to multilayered graphene oxide (MLGO), we tested whether three different U(VI):MLGO ratios (3 ppm U; 20−210 mg L−1 MLGO) affect the ability of nonelectrostatic and diffuse layer models to predict U(VI) adsorption behaviors across a range of ionic strength (1−100 mM) and pH (2−9.5) conditions. Model formulations assumed interactions between discrete MLGO surfaces sites and the most abundant aqueous U(VI) complex(es) within a given pH range. We determined that the observed extents of U(VI) binding require adsorption of more than one U(VI) species (UO22+ and uranyl hydroxide(s) and/ or carbonate(s)) and calculated the respective stability constants for the important U(VI)-MLGO surface complexes. The results also unequivocally illustrated that models using adsorption data from treatments with higher U(VI):MLGO ratios provide better fits throughout the tested range of experimental conditions, meaning that the U(VI)-MLGO stability constants calculated herein can be confidently applied to a range of natural or engineered systems.
■
INTRODUCTION The development of nuclear technologies has resulted in significant legacy concentrations of radionuclides in some soil and groundwater systems, including those near former nuclear weapons research/test facilities1−3 and abandoned mine tailings.4−6 In aerobic environments, U is present in its oxidized form, U(VI), which has a complex aqueous speciation that depends on the surrounding solution pH and composition. For example, at relatively low pH, U(VI) exists as the divalent uranyl ion (UO22+), and in systems containing low Ca concentrations, its speciation transitions to aqueous complexes that exhibit weak positive charges at circumneutral pH and strong negative charges at higher pH. In general terms, these aqueous U(VI) complexes include uranyl hydroxides and uranyl carbonates, respectively. The UO22+ ion readily associates with the negatively charged surfaces common in soil and groundwater systems,7−9 while the uranyl complexes present at higher pH values are often more poorly adsorbed.10,11 Hence, the fate and mobility of U(VI) in contaminated environments is greatly influenced by its speciation and the tendencies for these distinct species to adsorb to stationary-phase materials. © XXXX American Chemical Society
In fact, by purposefully harnessing these reactions, U(VI) can be captured or immobilized by engineered material adsorbents, which has become a widespread approach for the remediation of U(VI) contaminated environments.12−18 Multiple researchers have reported that graphene oxide (GO) nanosheets produced from the chemical oxidation and exfoliation of graphite exhibit a particularly striking affinity to adsorb U(VI),19−22 making it an outstanding candidate for further development as a U(VI) remediation tool. Both Li et al.19 and Zhao et al.20 separately reported maximum U(VI) adsorption capacities for GO of between ∼100 and ∼300 mg g−1 (pH 4 or 5; T = 293 K), which at the time of their publications, were the highest values yet reported for any material adsorbent. However, these studies used Langmuir−Freundlich isotherms to model U(VI)-GO interactions, and as a result, do not provide molecular-scale insights on the possible identities of the aqueous uranyl Received: November 15, 2016 Revised: June 7, 2017 Accepted: June 19, 2017
A
DOI: 10.1021/acs.est.6b05776 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
that result contained a mixture of particle morphologies. Relative to single- or few-layered GO, multilayered GO (MLGO) particles are more easily separated from contaminated solutions via filtration or sedimentation, making them more practical for use in engineered water treatment and remediation systems. As a result, this study focused exclusively on MLGO particle morphologies. MLGO particles were isolated from the other particle morphologies by repeatedly sedimenting particle suspensions under centrifugation (4000g for 5 min) and decanting the supernatant, which largely removed the single- and fewlayered nanosheets that remain suspended during the centrifugation process. The concentrations of the final stock MLGO suspensions averaged ∼2 g L−1, as determined by the difference between the wet and dry (105 °C for >24 h) mass of 20 mL aliquots for each final suspension. This stock suspension was diluted as needed to achieve the target MLGO concentrations described below. We utilized atomic force microscopy (AFM; XE-70, Park Systems, Santa Clara, CA) to characterize the physical properties of the MLGO particles used in subsequent U(VI) adsorption experiments. To this end, an aliquot of the MLGO stock suspension was diluted 1:4 with ethanol and deposited on freshly cleaved mica. The AFM was operated in noncontact mode and images were flattened using the WSxM software package from Nanotec Electronica S.L. (Madrid, Spain). The AFM analyses resulted in determination of z-dimension, or height, which we used to infer the extent of exfoliation present in the MLGO suspensions. Batch U(VI)-MLGO Adsorption Studies. MLGO suspensions of varying concentrations (20, 40, or 210 mg L−1) and initial ionic strength (1 mM, 10 mM, and 100 mM via the addition of NaClO4) were evaluated for their ability to adsorb a fixed concentration of U(VI) (3 ppm; 1.26 × 10−5 M) in aerobic conditions between approximately pH 2.0 and pH 9.5. We conducted two replicate batches for each treatment, with each batch consisting of multiple 5 mL samples equilibrated to different pH values using 0.1 to 1.0 M HNO3 or NaOH. Final U(VI) concentrations were achieved by diluting a 1000 ppm (4.20 × 10−3 M) inductively coupled plasma (ICP) U(VI) standard solution ((UO2)(NO3)2 in 2% HNO3) and neutralizing the resulting solution immediately before sample preparation. Exposures between dissolved U(VI) and the MLGO particles occurred for at least 4 h, as kinetic experiments associated with similar batch metal adsorption studies indicated a steady state occurs within this length of time. After the equilibration period, a final pH was measured, and the MLGO particles were removed from suspension by centrifugation (4000g for 5 min) followed by filtration (0.2 μm PTFE syringe filters) of the supernatant. The resulting solutions were acidified and analyzed via ICP-optical emission spectroscopy (Optima 2000DV, PerkinElmer, Waltham, MA) against matrix-matched aqueous U(VI) concentration standards. Chemical Equilibrium Modeling. Surface complexation approaches have been used successfully to model metal adsorption onto mineral surfaces,28 and this approach eliminates many of the problems associated with the partitioning approaches described above. However, surface complexation models require a molecular-scale understanding not only of the surfaces involved, but also of the adsorption/desorption mechanisms. A surface complexation model treats the adsorbed metal as a species whose stability can be quantified with an equilibrium constant for the adsorption reaction. If the equilibrium constant values for each of the important equilibria
complexes being adsorbed and cannot be used to calculate the intrinsic stability constants for the associated U(VI)-GO surface complexes. Unlike the empirical approaches utilized in the studies described above, three additional adsorption studies have used a more mechanistic surface complexation modeling approach to characterize U(VI)-GO binding extents.23−25 Sun et al.23 assumed an interaction between amphoteric surface hydroxyl groups and the UO22+ ion to model their observed adsorption behaviors. The modeling approach of Hu et al.25 was similar, but assumed a bidentate binding mechanism for an additional uranyl hydroxide species as pH increased. Xie et al.24 utilized a combination of carboxyl and sulfonate functional groups as UO22+ adsorption sites, focusing primarily on U(VI) adsorption at low pH. Each study used a single experimental U(VI):GO ratio, which in all cases resulted in adsorbed U(VI) extents reaching ∼100% while still at acidic pH values where the UO22+ ion dominates U(VI) speciation. Adsorption extents remained at very high levels across the remainder of their respective experimental pH ranges. In this study, we hypothesize that very high experimental adsorption extents (i.e., those at or near 100% over a wide range of pH) are not ideal for surface complexation models, as they result in calculated stability constants that are poorly constrained and, similar to empirical adsorption approaches, only apply to the tested conditions. Adsorption extents are not measured directly and instead calculated as the difference between the known initial dissolved metal concentration and that which remains in solution after exposure to the adsorbent. Hence, when a high percentage of the metal is adsorbed, the concentration of metal measured in solution is generally at or near the instrument detection limit, meaning that even very small absolute uncertainties in these measurements can have a dramatic impact on the stability constant calculations. In addition, in systems that exhibit complete removal of metal from the aqueous phase, it is impossible to determine whether the adsorbent could have captured more if a higher initial dissolved metal concentration had been provided. For these reasons, targeting adsorbate:adsorbent ratios that result in more equally matched concentrations of aqueous and adsorbed metals is necessary in order to provide rigorous constraints on the identity and thermodynamic stability of the important surface complexes. We tested this hypothesis by conducting bulk U(VI) adsorption experiments at three different U(VI):GO ratios under a wide range of environmentally relevant ionic strength (1−100 mM) and pH (2−9.5) conditions. We used these experimental results to construct surface complexation models that yielded discrete U(VI)-GO stability constants for each experimental treatment, which were then tested for their abilities to predict the U(VI) adsorption behaviors observed across the entire range of experimental conditions. The results unequivocally illustrated that models that use adsorption data from treatments with higher adsorbate:adsorbent ratios provide better fits to the totality of experimental data. Hence, the U(VI)-GO stability constants calculated in this study can be confidently applied to an array of natural or engineered systems.
■
EXPERIMENTAL PROCEDURES MLGO Preparation and Characterization. GO particles were synthesized using the method of Hummers and Offeman,26 which has been summarized elsewhere27 and in the Supporting Information (SI) Section S1. The GO suspensions B
DOI: 10.1021/acs.est.6b05776 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
suspensions used in this study. Our AFM analyses indicated that MLGO suspensions contained particles ranging in height up to approximately 10 nm and consisting of layered structures of individual nanosheets. For comparison, the height of a single hydrated GO nanosheet is roughly 1 nm.33−35 MLGO-Free U(VI) Speciation Modeling. To facilitate the interpretation of the U(VI)-MLGO adsorption data, we conducted MLGO-free U(VI) speciation modeling to determine the identity of the aqueous uranyl complexes present under each tested experimental condition. Those aqueous complexes that exhibited greater than 5% abundance in the speciation modeling are illustrated in Figure 1 as a function of
in a system are known, then the distribution of metals between various reservoirs (i.e., in solution or adsorbed onto surfaces) can be calculated. We tested several different surface complexation modeling approaches, including various formulations of a nonelectrostatic model (NEM) and a diffuse layer model (DLM), for their abilities to account for the observed U(VI)-MLGO adsorption behaviors across the tested range of pH, ionic strength, and MLGO concentration. All surface complexation modeling was completed using FITEQL 2.0,29 and accounted for the formation of aqueous U(VI) complexes using the stability constants in SI Table S1.30,31 Proton-active surface site concentrations and acidity constants for the NEM (SI Table S2) and DLM (SI Table S3) formulations were obtainted from Duster et al.,27 who calculated these values from potentiometric titrations under experimental conditions that were similar to those of our batch U(VI) adsorption experiments (e.g., identical background electrolyte and similar pH and ionic strength ranges). Unlike the amphoteric surface sites used in some previous GO surface complexation models,23,25 our models assumed that MLGO surface sites are not capable of being doubly protonated, which is consistent with findings that nearsurface charge values (i.e., ζ potential) remain negative throughout our tested range of experimental pH.23 We always assumed a 1:1 stoichiometry between discrete MLGO surface sites and adsorbed U(VI) species because the X-ray absorption spectroscopy (XAS) measurements we previously performed on similar systems implicated monodentate binding as the most likely adsorption mechanism.32 However, these same XAS measurements could not define the stoichiometries of the adsorbed U(VI) species because the number/identity of deprotonated MLGO surface sites and the speciation of the adsorbed U(VI) both change with pH. In addition, using XAS, it is extremely difficult (if not impossible) to distinguish between the C atom from the MLGO structure and the C atom from the uranyl carbonates that are likely to represent one or more of the adsorbed U species (as discussed later) because each C atom could presumably exist in the second shell around U. Consequently, we chose to assume model reactions between the most abundant aqueous U(VI) species present within a given pH range and MLGO surface sites that are deprotonated in the same pH range. Our assumed model reactions represent a probable conceptual model because in order for a less abundant U(VI) species to be selectively adsorbed over one of greater abundance, the calculated stability constant would have to be uncharacteristically large. Additional details of our assumed reactions are provided in subsequent sections. To establish these species abundances, we conducted U(VI) speciation modeling for MLGO-free systems using FITEQL 2.0 following the same basic approach as described above. Stability constants for the aqueous uranyl complexes in these models are found in SI Table S1 and all models exhibit the same pH and background electrolyte conditions as those from the batch adsorption experiments. We assume that the precipitation of solid U(VI) species is negligible in our systems, as our previous XAS measurements32 on U-MLGO systems found none of the U−U bonds that would be indicative of U precipitation, even at U concentrations that are significantly higher than the concentration used here.
Figure 1. Abundance of U(VI) species for MLGO-free systems in equilibrium with atmosphere across the indicated ranges of pH and ionic strength. Boxed species are those used in subsequent modeling efforts. Gray boxes highlight pH ranges where the dominant U(VI) species changes. Only species that exceed 5% abundance are shown. Abundances were calculated using FITEQL 2.0 and the aqueous stability constants provided in SI Table S1. ((U)total = 1.3 × 10−5 M).
pH and ionic strength. There were relatively small but significant differences in the U(VI) speciation as a function of ionic strength, due to the influence of ionic strength on the calculated activity coefficients for the (highly) charged uranyl complexes, especially under high pH conditions. At the median ionic strength evaluated (i.e., 10 mM), the UO22+ ion dominated these systems below about pH 5.2, while the 3:5 uranyl:hydroxide complex ((UO2)3(OH)5+)) became the most abundant species between pH 5.2 and pH 6.3. Several other uranyl hydroxides were also present in relatively low abundance in this pH range. Above pH 6.3, the U(VI) speciation was dominated by negatively charged uranyl complexes, including the 2:1:3 uranyl:carbonate:hydroxide ((UO2)2(CO3)(OH)3−) between pH 6.3 and pH 8.2, and the 1:3 uranyl carbonate ((UO2)(CO3)34−) above pH 8.2. This speciation modeling informed our assumptions regarding adsorbed complexes at MLGO particle surfaces, both in the qualitative discussion below and the surface complexation modeling conducted in later sections. Batch U(VI) Adsorption Experiments. The results of the batch U(VI) adsorption experiments are depicted in Figure 2 as a function of pH, ionic strength, and MLGO concentration. U(VI) adsorption curves for each individual treatment are comprised of the results from two replicate experimental batches, which largely overlapped and are therefore not distinguished. The term treatment is used herein to describe a specific
■
RESULTS AND DISCUSSION MLGO Nanosheet Characterization. SI Figure S1 depicts a representative AFM image of MLGO particles from the stock C
DOI: 10.1021/acs.est.6b05776 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
(>95%) adsorption over approximately 1.5 and 4 pH units, respectively, the adsorption extents exhibited by the 20 mg L−1 MLGO treatment peaked at approximately 90% over a very narrow pH range. The influence of ionic strength on adsorption behaviors in those systems containing the largest concentrations of MLGO was largely obscured by the high extents of adsorption. At the lowest tested MLGO concentration (20 mg L−1 MLGO), ionic strength influenced the extent of U(VI) adsorption at the MLGO surface, but the extent of this change within a range of ionic strengths that span over two orders-of-magnitude was relatively small. These observations are consistent with U(VI)-GO adsorption20 and XAS32,36 studies in the literature that suggest inner-sphere complexation for U(VI) at the GO surface. The adsorption of U(VI) increased with increasing pH through about pH 4 for the 210 mg L−1 MLGO treatment and about pH 6 for the 20 mg L−1 MLGO treatment. Duster et al.27 previously used an NEM to calculate pKa values of 4.35, 6.32, 8.32, and 9.77 (SI Table S2) for four distinct MLGO surface sites, likely including carboxyl and hydroxyl groups that exist on the basal planes or at the edges of the MLGO particles (see Duster et al.27 for more details). Hence, at very low pH, most surface sites are protonated and consequently we measured relatively low extents of U(VI) adsorption below pH 3.5. As pH increased, MLGO surface functional groups were increasingly deprotonated to become negatively charged and we attributed the dramatic increase in adsorption through about pH 5.2 in each treatment to the binding of the positively charged UO22+ ion that dominated U(VI) speciation in this pH region (Figure 1). The appreciable extents of U(VI) adsorption above pH 5.2 for each experimental treatment were likely a result of the adsorption of other aqueous U(VI) complexes. Positively charged complexes continue to dominate the U(VI) speciation in these experimental systems up to approximately pH 6.3, which may explain the continuing high extents of U(VI) adsorption at these pH values. The transition to negatively charged U(VI) species at higher pH values diminished (though did not eliminate) adsorption onto the also negatively charged MLGO surface, especially for the 20 mg L−1 and 40 mg L−1 MLGO suspensions where adsorption sites were limited. We use surface complexation modeling in later sections to provide more insights on these reactions. The greater number of total available adsorption sites in the 210 mg L−1 MLGO experiments partially compensated for the lower affinity between the U(VI) species and the MLGO surface functional groups that we observed in the 20 mg L−1 and 40 mg L−1 experiments at high pH. In fact, regardless of the ionic strength, the 210 mg L−1 MLGO experiments exhibited complete adsorption of U(VI) between about pH 4 and pH 7.5, with adsorption gradually decreasing but remaining above 80% through pH 9.5. The measured extents of U(VI) adsorption at high pH for these treatments were significantly higher than those measured in previous studies for hematite11 or ferrihydrite,10 each of which exhibited a precipitous decrease in U(VI) adsorption above approximately pH 8.7. While it is difficult to compare calculated material site concentrations from these studies due to differences in analytical approaches, the greater affinity for U(VI) displayed by the 210 mg L−1 MLGO treatments at high pH occurred despite having many-fold lower site:U(VI) molar concentration ratios relative to those found in the iron oxide-bearing experiments. This comparison suggests that MLGO particles may be successfully utilized for U remediation, treatment, and/or capture in particularly challenging
Figure 2. Adsorption of U(VI) (3 ppm; 1.3 × 10−5 M) by MLGO across the indicated ranges of pH, ionic strength, and MLGO concentration.
experimental combination of MLGO concentration and ionic strength, with each of the nine treatments depicted in Figure 2 considered separately, where appropriate, for subsequent discussions and modeling efforts. MLGO particles exhibited a significant ability to adsorb U(VI) throughout a wide range of experimental conditions. The MLGO concentration played a large role in determining the extent of U(VI) adsorbed, with the more concentrated MLGO suspensions adsorbing higher percentages of U(VI) over a broader range of pH due to the greater number of total reactive surface sites available. In fact, while the 40 mg L−1 and 210 mg L−1 MLGO experiments achieved near-complete D
DOI: 10.1021/acs.est.6b05776 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology environments, including those exhibiting relatively high pH and ionic strengths. Surface Complexation Modeling. One goal of this study was to construct a single surface complexation model that best fits the observed U(VI) adsorption data across the wide tested ranges of pH, ionic strength, and MLGO concentration. We began by building various NEM and DLM formulations and then used these models to calculate stability constants from the adsorption data associated with each of the nine experimental treatments depicted in Figure 2. The specific models that were tested are described below. SI Table S4−S6 and Table S7−S11 provide the calculated treatment-specific stability constants for each tested NEM and DLM formulation, respectively, while a summary of the calculated stability constants for the NEM formulations is provided in Table 1. We then tested the treatmentspecific stability constants averaged across ionic strength for their ability to predict the adsorption behaviors under the other experimental conditions of this study. If a given model formulation accurately accounted for the electrostatic conditions at the MLGO surface, then the treatment-specific stability constants exhibited a relatively narrow range when considered across ionic strength and the corresponding average represented the best chance at fitting adsorption data from each of the nine experimental treatments. The best-fitting NEM and DLM were taken to be those that yielded the smallest residual differences between measured and modeled values over a range of pH rand ionic strength, as determined by the V(Y) goodness-of-fit parameter output from FITEQL29 and visual fits to the range of experimental adsorption data. The NEM and DLM differ in their assumption regarding the electrostatic conditions at the surface of a material, with the DLM being more capable of accounting for large ionic strength dependencies in experimental adsorption data. Because we observed only a small ionic strength dependency in the U(VI) adsorption behavior, the NEM and DLM formulations were equally capable of fitting the experimental data and the conclusions that we draw from the analysis of one model type applies equally well to the other. The NEM is the simpler model and requires fewer adjustable parameters than the DLM (e.g., the DLM requires a MLGO surface area parameter that is particularly difficult to measure37), and consequently, we opted to focus primarily on the NEM results in the main text. However, we will show evidence for the similarity between the NEM and DLM results in subsequent discussions. Measurement of the total adsorbed U and pH, in conjunction with known site concentrations and acidity constants (SI Table S2) and equilibrium constants for the aqueous complexation reactions (SI Table S1), enable us to solve for the stability constants of the important U-MLGO surface complexes and their concentrations at each condition studied. At relatively low pH (