water interface using the Charge Distribution Multi-Site Complexation (CD-MUSIC) model

water interface using the Charge Distribution Multi-Site Complexation (CD-MUSIC) model

Journal of Colloid and Interface Science 418 (2014) 147–161 Contents lists available at ScienceDirect Journal of Colloid and Interface Science www.e...

2MB Sizes 1 Downloads 46 Views

Journal of Colloid and Interface Science 418 (2014) 147–161

Contents lists available at ScienceDirect

Journal of Colloid and Interface Science www.elsevier.com/locate/jcis

Surface complexation modeling of Hg(II) adsorption at the goethite/ water interface using the Charge Distribution Multi-Site Complexation (CD-MUSIC) model Jeremiah E. Mangold 1, Chang Min Park 2, Howard M. Liljestrand, Lynn E. Katz ⇑ The University of Texas at Austin, Department of Civil, Architectural and Environmental Engineering, 301 E Dean Keeton Street Stop C1786, Austin, TX 78712-1173, USA

a r t i c l e

i n f o

Article history: Received 20 September 2013 Accepted 31 October 2013 Available online 9 November 2013 Keywords: Adsorption Mercury Chloride Carbonate Goethite (a-FeOOH) CD-MUSIC model

a b s t r a c t Hypothesis: Previous attempts to describe Hg(II) adsorption onto mineral surfaces using surface complexation models (SCMs) have proven unsuccessful and/or require the use of hypothetical surface species. Given that metal ion adsorption at the mineral–water interface is greatly influenced by mineral surface heterogeneity and the presence of competing adsorbates in solution, it stands to reason that estimating the crystal face composition (CFC) of the mineral surface and the extent of carbonate contamination in the experimental system will improve SCM predictions. Experiments: The Charge Distribution Multi-Site Complexation (CD-MUSIC) model was used to simulate experimental Hg(II) adsorption data, collected on the iron hydroxide mineral goethite, in the presence and absence of competing adsorbates and complexing ligands as a function of pH and ionic strength. The CFC of each goethite sample studied was predicted using a newly discovered relationship between goethite’s proton reactive site density (NH) and specific surface area (SSA). Carbonate’s presence in the experimental systems was determined utilizing a novel methodology developed in this work. Findings: The CD-MUSIC model developed in this study accurately predicted Hg(II) adsorption onto goethite over the entire range of experimental conditions investigated while only employing surface species consistent with spectroscopic evidence. Ó 2014 Published by Elsevier Inc.

1. Introduction Adsorption of ions to mineral phases in solution is of great importance in the regulation of many trace metals in the environment. Unlike other trace metals such as Cd, Zn, and Cu, Hg(II) has unique physicochemical properties especially with respect to its redox speciation and ability to form stronger complexes with carbonate and chloride. Mercury contamination has occurred extensively in a number of industries; from its historical use in cosmetics, pharmaceutics, and the paper industry to its modern Abbreviations: CD-MUSIC model, Charge Distribution Multi-Site Complexation 2 model; CFC, crystal face composition; h iNS, site density (sites/nm ); NH, proton reactive site density (sites/nm2); CO2 , all carbonate present in the solid – 3 TOT solution system from both surface and aqueous species; EXAFS spectroscopy, extended X-ray absorption fine structure spectroscopy; VIL70, Goethite sample used by Villalobos and Leckie [32,33]; BAR76, Goethite sample used by Barrow and Cox [5]; BON15, Goethite sample used by Bonnissel-Gissinger et al. [6]. ⇑ Corresponding author. Fax: +1 (512) 471 5870. E-mail addresses: [email protected] (J.E. Mangold), changmin.p@samsung. com (C.M. Park), [email protected] (H.M. Liljestrand), lynnkatz@mail. utexas.edu (L.E. Katz). 1 Rich Lab, 342 Computer Court, Anderson, SC 29625, USA. 2 Cheil Industries Inc., 332-2, Gocheon-dong, Uiwang-si, Gyeonggi-do 437-711, Republic of Korea. 0021-9797/$ - see front matter Ó 2014 Published by Elsevier Inc. http://dx.doi.org/10.1016/j.jcis.2013.10.066

usage in chlor alkali plants (45% of all domestic mercury emissions), writing devices and switches, lighting, and dental work [1]. The fate of Hg(II) in the environment is often controlled by its adsorption behavior on fixed or mobile adsorbents [2]. Among the most common inorganic adsorbents including hydroxides, oxyhydroxides, and clay minerals; metal (hydr)oxides have the highest affinity for inorganic ions in solution due to their high specific surface area (SSA), surface charge, and reactive site density [3]. While the adsorption behavior of Hg(II) on metal (hydr)oxides has not been extensively studied, it is understood that both aqueous complexation and potential ternary complex formation can impact the extent of adsorption to mineral surfaces [4–8]. Numerous experimental and modeling studies for metal ion adsorption have been performed and the data obtained has been fit to a variety of adsorption isotherm or surface complexation models (SCMs) [3,5,6,8–19]. Surface complexation models that allow formation of inner- and outer-sphere complexes between surface hydroxyl groups and complexing aqueous ions offer the greatest potential for describing Hg(II) adsorption because they have the ability to account for the impacts of pH, ionic strength, and ternary complex formation. There are a number of surface complexation models that differ in their description of the interfacial region both with respect to the description of the relationship

148

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161

between surface charge and potential, and the location of sorbing species [20,21]. The Non Electrostatic Model (NEM) and the Diffuse Layer Model (DLM) offer simplistic descriptions of the double layer region that require minimal parameter estimation but only allow formation of inner-sphere complexes. Both the Triple Layer Model (TLM) and the Charge Distribution Multi-Site Complexation (CD-MUSIC) model provide a more complete description of the interfacial region and the means to account for both inner- and outer-sphere complexation as well as ligand exchange [22] that is consistent with spectroscopic measurements. The CD-MUSIC model can be differentiated from the TLM by three major characteristics; representation of surface acidity, placement of ions and charge in electrostatic planes, and representation of reactive surface adsorptions sites [23–26]. Both models have potential for describing Hg(II) adsorption in complex systems that include complexing ligands such as Cl. The majority of macroscopic Hg(II) adsorption studies were conducted by uptake measurements on numerous natural or synthetic substrates as a function of pH [5,6,9,11,13,27–30]. MacNaughton and James [11] investigated the adsorption of Hg (II) on SiO2 and showed a sharp increase in the extent of adsorption in the range of pH 3 to 4 with a maxima at pH 4. When chloride was present in solution, Hg(II) removal was still attributed to Hg(OH)2 adsorption. Lockwood and Chen [10] investigated Hg (II) adsorption on Fe(OH)3 and also showed a sharp increase in adsorption in the range of pH 3 to 4 and Cl inhibition of Hg(II) adsorption at neutral pH. They also concluded that the hydrolysis products HgOH+ and Hg(OH)2 were the major species controlling Hg(II) adsorption. Several previous investigators have applied surface complexation/precipitation modeling to describe this adsorption behavior for Hg(II) onto silica and iron oxides as a function of pH and Cl concentrations. Tiffreau et al. [13] employed the Diffuse Layer Model (DLM) to describe adsorption onto a-SiO2 (quartz) and HFO (amorphous hydrous ferric oxide). While the DLM incorporated ternary surface complexes such as „SAOHgOH and „SAOHgCl and required fewer parameters to describe adsorption compared to the TLM or CD-MUSIC, the DLM overpredicted Hg(II) adsorption in a number of simulations using data from MacNaughton [31] and Kinniburgh and Jackson [9]. All of these modeling

efforts were conducted in the absence of spectroscopic data to guide the selection of surface complexes. In this study, a CD-MUSIC model describing Hg(II) adsorption and aqueous speciation has been developed over a range of experimental conditions to provide a tool for predicting the fate and transport of mercury down gradient of a spill or after residual Hg(II) remains at a remediated site. Competitive Hg(II) adsorption on ubiquitous soil minerals such as goethite has been evaluated over a range of experimental conditions. Previous research efforts utilizing the classical 2-pK, CCM, and TLM models were unable to predict Hg(II) adsorption adequately or did not employ surface species that are consistent with molecular scale analyses. In addition, these studies did not account for the presence of carbonate in their experimental systems despite telltale signs of carbonate contamination [5,6]. Thus, the development of a more accurate surface complexation model that incorporates carbonate adsorption and utilizes spectroscopic data to guide the selection of surface species for Hg(II) adsorption onto goethite is warranted.

2. Materials and methods 2.1. Experimental data employed Several sets of adsorption data were obtained from the literature for this modeling study. Data for carbonate adsorption onto goethite in an open system at atmospheric CO2 values were obtained from Villalobos and Leckie [32,33] to account for the undesired carbonate contamination present in most of the published data for Hg(II) adsorption onto goethite. The reported specific surface area of this goethite was 70 (±2) m2/g determined by N2 BET (Brunauer, Emmett and Teller) adsorption on the dry (105 °C) material. The influence of pH, mercury surface loading, and chloride concentration on Hg(II)’s adsorption to goethite was evaluated using data from Barrow and Cox [5]. In their research, equilibrium studies were conducted over a range of conditions: nitric acid or sodium hydroxide was used to vary pH, mercury(II) nitrate was used as the mercury source, sodium chloride was used as the

Fig. 1. Relationship between SSA (m2/g) and proton-reactive site Ns (sites/nm2) for the different goethite systems. Symbols denote data taken from Salazar-Camacho and Villalobos [48]. Solid line denotes a linear trend line of the data points.

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161

chloride source, and sodium nitrate was employed to maintain constant ionic strength. The solid concentration for each adsorption experiment was not reported but could be approximated via a mass balance approach utilizing the total, aqueous, and sorbed concentrations of Hg(II) reported. The concentration of goethite used was found to range between 4.02 and 4.51 g/L. The specific surface area was determined to be 76 m2/g using ethylene glycol monoethyl ether. The experimental description did not indicate that carbonate was eliminated from the system. Bonnissel-Gissinger et al. [6] investigated Hg(II) adsorption onto a commercial goethite (Bayferrox 910, standard 86) from Bayer. The specific surface area of 15 m2/g for goethite was obtained by N2 BET analysis. The pHpzc of 7.85 and the surface site density of 5.5 OH nm2 were determined and confirmed by potentiometric titrations [34]. Sodium nitrate and sodium chloride were used as the background electrolytes and a solid concentration of 10 g/L was utilized in all experiments. The pH of the solution was adjusted using nitric acid and de-carbonated sodium hydroxide. The Hg(II) concentration of 12.4 lM was prepared from a Hg(NO3)2 standard solution. The physicochemical parameters and suspension properties of the three systems studied in this work are summarized in Table 1. For the three goethite systems studied, all adsorption experiments were conducted at 25 °C. Proton reactive site density is calculated using Eq. (3) and discussed in detail below. Potentiometric acid base titration data collected at several different ionic strengths gives rise to a typical surface charge profile such as that shown in Fig. 2. The crossover point or intersection

Table 1 The physicochemical parameters and suspension properties used in model simulations. Parameter

Specific surface area (m2/g) Solid’s concentration (g/L) pHpzc Proton reactive site density (sites/nm2) a b c d e

Goethite sample VIL70a

BAR76b

BON15c

70 13.8d and 9.4e 8.9 6.97

76 4.02–4.51 7.7 6.81

15 10 7.85 8.51

Villalobos and Leckie [32,33]. Barrow and Cox [5]. Bonnissel-Gissinger et al. [6]. Used in acid–base titration in 0.1 and 0.015 M NaNO3. Used in carbonate adsorption on goethite.

149

of the data collected at different ionic strengths is termed the point of zero salt effect and is an estimate of the pHpzc [35]. Previous research has suggested that the titration data and pHpzc is a function of the solid synthesis method and the nature of electrolyte used [16]. However, a number of carefully controlled potentiometric titration experiments conducted for goethite have obtained similar estimates of the pHpzc. For example, the pHpzc for goethite has been determined to be 9.2 in NaCl, 9.0 in NaNO3 [32], 9.1 in NaCl [36], 8.9 [37], 9.0 ± 0.3 [38] in suspensions of NaClO4, and 9.2–9.3 in NaNO3 [39]. In contrast, the surface charge data obtained from Barrow and Cox [5] and Bonnissel-Gissinger et al. [6] exhibited a low pHpzc of approximately 7.7 and 7.85, respectively. Low estimates of the pHpzc are often attributed to the adsorption of carbonate species onto goethite or contamination of the suspension with carbonate [32,36,38,40–43]. Given this information, it was concluded that the mercury adsorption studies of Barrow and Cox [5] and Bonnissel-Gissinger et al. [6] both had carbonate present in the system. 2.2. Modeling approach (CD-MUSIC Model) In solid–solution systems, ion adsorption behavior is greatly influenced by the crystal morphology of the mineral surface [44– 48]. The crystal faces that make up the mineral surface possess reactive surface oxygens, herein referred to as surface sites, which can vary in type and density (sites/nm2) from one crystal face to another. There are four principal crystal faces found on goethite: the (1 0 1), (0 0 1), (2 1 0), and (0 1 0), using the Pnma space group [23,40,47,49–53]. The surface site types present on these four crystal faces are presented in Table 2 along with their reactive site densities (NS). Each surface site type is thought to possess its own unique reactivity to adsorbing ions, including protons [23,25,52,54]. Doubly coordinated sites located on the (1 0 1) and (0 0 1) crystal faces of goethite are considered to be non-reactive [23,25,40,47,52], and only one third of the triply coordinated sites present on the (1 0 1) and (0 0 1) crystal faces are regarded as reactive [25,26,47,49,55,56]. Doubly coordinated sites located on the (101) and (00 1) crystal faces of goethite are considered to be non-reactive [23,25,40,47,52]. Only one third of the triply coordinated sites present on the (10 1) and (00 1) crystal faces are considered reactive (3.03 and 3.34 Fe3OI sites/nm2, respectively) [25,26,47,49,55,56]. Proton reactive site density (NH) is the sum of reactive singly and triply coordinated surface sites [23,25,46,47,52,54,57]. The CD-MUSIC model developed by Hiemstra et al. [23,44,45] and Hiemstra and Van Riemsdijk [49] describes surface acidity

Fig. 2. Potentiometric titration data for goethite sample VIL70 at two different ionic strengths using NaNO3 as the background electrolyte. Symbols denote experimental data taken from Villalobos and Leckie [32,33] and lines denote CD-MUSIC model simulations using the (a) segregate and (b) composite approaches to describing the mineral’s surface (see text for more details). Reactions listed in Table 3 were used in performing model simulations.

150

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161

Table 2 Reactive site densities of singly, doubly, and triply coordinated sites on the predominant crystal faces of goethite. Site type

BFeOH BFe2OH BFe3O NH a b c

Crystal face Ns (sites/nm2) (1 0 1)a

(0 0 1)b

(2 1 0)a

(0 1 0)c

3.03 0 3.03 6.06

3.34 0 3.34 6.68

7.5 7.5 0 7.5

9.1 9.1 0 9.1

Venema et al. [54]. Gaboriaud and Ehrhardt [40]. Lutzenkirchen et al. [51].

using a 1  pK assumption for singly and triply coordinated surface sites on goethite (Eqs. (1) and (2), respectively).

BFeOH0:5 þ Hþ $ BFeOHþ0:5 2

ð1Þ

BFe3 OH0:5 þ Hþ $ BFe3 OHþ0:5

ð2Þ

Doubly coordinated surface sites present on goethite are considered to be non-reactive to protons within the normal pH range tested [23,25,47,52,54]. This assumption is based on the findings of Hiemstra et al. [23] who predicted the log KH values (i.e., protonation constants) for goethite’s different surface sites via a linear relationship the authors discovered between a specie’s log KH value and its oxygen atom’s formal charge. Using the bond valence approach presented in Brown and Altermatt [58], Hiemstra et al. [49] was able to calculate the formal charge of the „Fe2OH surface oxygen and subsequently predict the surface site’s log KH value to be 0.4, too low for „Fe2OH to react with protons in the normal pH range considered. Detailed descriptions of the CD-MUSIC model, in particular its depiction of the mineral–water interface, as well as further explanation of site reactivity and the charge distribution factor, can be found elsewhere (e.g., [23,49,54,59,60]). Molecular scale analyses provide researchers with a more accurate understanding of the adsorptive mechanisms/processes occurring at the mineral surface. Extended X-ray absorption fine structure (EXAFS) spectroscopy combined with molecular modeling (e.g. Cerius2 from Accelrys, Inc. and PC Spartan Pro from Wavefunction, Inc.) provides a nondestructive in situ analytical method uniquely suited to determine a the identify of nearest neighbors of a particular metal ion, number of nearest neighbors, interatomic distances, and degree of structural order associated with metal ion surface complexes. In addition, several studies have used bond valence analysis that assesses the stability of different potential sorption complexes at the surface by placing basic bonding constraints on various modes of metal ion sorption [61–64]. Literature data from molecular scale analyses were reviewed to select the predominant adsorptive surface species for Hg(II), CO2 3 , and HgACl ternary complexes. Hg(II)’s adsorption onto goethite was described using the CDMUSIC model to depict the underlying physicochemical processes occurring at the solid–solution interface via FITEQLC [65]. The surface and aqueous complexation reactions were formulated in the preprocessor of the existing framework of FITEQL 4.0 [66]. The entire modeling approach applied in this paper was based on a sequential optimization approach: (1) simulation of the potentiometric titration data for VIL70 to determine the intrinsic surface protonation and ion pair formation equilibrium constants; (2) simulation of carbonate adsorption on VIL70 with surface species proposed by Hiemstra et al. [67] and Rahnemaie et al. [68] to account for the low pHpzc reported by Barrow and Cox [5]; (3) simulation of the potentiometric titration data obtained from Barrow and Cox [5] assuming a fixed total carbonate concentration ð½CO2 3 TOT Þ and

using the affinity constants and parameters previously determined in steps 1 and 2; (4) simulation of Hg(II) adsorption on BAR76 using the data published by Barrow and Cox [5] and incorporating carbonate species with the same parameters and values used in the previous steps; and (5) simulation of Hg(II) adsorption onto the BON15 goethite sample using the data published by Bonnissel-Gissinger et al. [6] for model verification. Since FITEQL requires input values of concentration rather than activity, the pH values for each data set were converted to a molar concentration using the chemical equilibrium software Visual MINTEQ 3.0 [69] and then input into FITEQL as such. In addition, Visual MINTEQ 3.0 was used to check for precipitation under all conditions studied in this work. It should be noted that no precipitation was predicted to occur under any of the conditions investigated. The model’s performance in predicting adsorption edge data was assessed by determining the weighted sum of squares divided by degrees of freedom (WSOS/ DF) for each data set, where values between 0.1 and 20 reflect acceptable model fits of the experimental data [66]. Input uncertainties were taken from previous research for metal ion adsorption to iron oxides [66]. 2.3. Estimation of model parameters 2.3.1. Capacitance values Model calibration or fitting of the titration data taken from VIL70 was used to determine the log Kin values for goethite’s surface protonation reactions (Eqs. (1) and (2)) and ion pair formation reactions (cf. Table 3). The capacitance values (C1 and C2) for each system (VIL70, BAR76, and BON15) were determined through fitting each goethite sample’s titration data with the model. In this study C1„C2 based upon the work of Hiemstra and Van Riemsdijk [60], who after re-evaluating the work of Rahnemaie et al. [70] in which the position of specific electrolyte ions at the interface was investigated, suggested that the value of C2 is very similar to C1. Sverjensky [71] arrived at a similar conclusion in deciding that the outer layer capacitance used in the triple layer model (TLM) should be changed from 0.2 F/m2, as previously defined for the TLM, and set equal to C1. Additionally, recent modeling studies with CD-MUSIC have followed this same suggestion of setting C2 equal to C1 [68,72–74]. Charge for surface species is based on the Pauling bond valence approach used by Hiemstra et al. [23]. Using this approach, the charge on „FeOH and „Fe3O is 0.5, while the charge for „FeOH2 and „Fe3OH is +0.5. Dz0 and Dz1 are the charge distribution coefficients that represent the change in charge on the 0 and 1 electrostatic planes that occurs as a result of the surface reaction. Capacitance values were determined through model fitting of potentiometric titration data for VIL70 and BAR76 and of Hg(II) adsorption data for BON15 since no potentiometric titration data was available. For both segregate and composite approaches C1@[email protected] F/m2 for VIL70 and C1@C2 = 1.0 F/m2 for BON15. For BAR76, C1@C2 = 1.65 and 1.70 F/m2 using the segregate and composite approaches, respectively. In a study conducted by Evans et al. [75], the authors found that goethite suspensions purged with N2 gas between 1 and 2 months time achieved a measured pHPZC of 9.2; whereas goethite samples purged with nitrogen for only 1–2 days possessed a lower pHPZC value due to carbonate’s continued presence in the system. A similar investigation by Zeltner and Anderson [38] found that increased purging of goethite suspensions by N2 gas for periods of up to 2 months could increase the measured pHPZC value of the sample from 8.1 to 9.0. With these findings in mind, Lumsdon and Evans [36] conducted a literature review of pHPZC values previously reported for goethite in ‘‘carbonate free’’ systems. The pHPZC values were found to range from 7.5 to 9.38; demonstrating the variability in purging time employed by researchers on their

151

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161 Table 3 Summary of reactions used in CD-MUSIC for the segregate and composite approaches to characterizing goethite’s surface. Adsorbate

Approacha

Crystal faceb

Surface complexation reaction

H+

Segregate; composite

(1 0 1), (0 1 0)

Segregate; composite

(1 0 1)

BFeOH + H+ M BFeOH2 BFeOH + H+ + NO 3 M BFeOH2_NO3 BFeOH + H+ + Cl M BFeOH2 _Cl + BFeOH + Na M BFeOH_Na BFe3O + H+ M BFe3OH BFe3O + H+ + NO 3 M BFe3OH_NO3 BFe3O + H+ + Cl M BFe3OH_Cl BFe3O + Na+ M BFe3O_Na

Composite



2(BFeOH) + CO23  + 2H+ M (BFeO)2–CO + 2H2O

0.62

0.62



18.9

Segregate

(0 1 0)

(BFeOH1/2) + (BFe2OH1/2) + 2H+ + CO23  M (BFeO)–CO–(BFe2O) + 2H2O

0.62

0.62

21.0



Hg2+

Composite



Segregate

(0 1 0)

2(BFeOH) + Hg2+ M (BFeO)2H–Hg + H+ 2(BFeOH) + Hg2+ M (BFeO)2Hg + 2H+ (BFeOH1/2) + (BFe2OH1/2) + Hg2+ M (BFeO)–Hg–(BFe2O) + 2H+

0.8 0.03 0.12

0.2 0.03 0.12

– – 2.3

3.7 5.0 –

Hg2+ and Cl

Segregate; composite

(0 1 0)

BFeOH + Hg2+ + Cl M BFeO–HgCl + H+

0.02

0.02

7.9

7.7

CO23 

Dz 0

Dz1

1 1 1 0 1 1 1 0

0 1 1 1 0 1 1 1

Log Kin Segregate

Composite

8.6 7.6 7.6 0.45 11.7 10.7 10.7 0.45

8.6 7.6 7.6 0.45 11.7 10.7 10.7 0.45

a

Column labeled ‘‘Approach’’ specifies which surface characterization approach (i.e., segregate or composite) the reactions listed in the corresponding row are applicable for. Column labeled ‘‘Crystal Face’’ denotes which crystal face or faces are involved in a particular reaction when using the segregate approach to characterize the mineral’s surface. In the segregate approach, the surface is pictured as being composed of different crystal faces that each possess a particular number of singly, doubly, and triply coordinated sites. The composite approach assumes all sites of a given type are identical and the mineral surface is pictured as being composed of only singly, doubly, and triply coordinated sites with no further differentiation; hence the ‘‘Crystal Face’’ column is not applicable for the CD-MUSIC model utilizing this approach. b

goethite suspensions prior to conducting adsorption experiments. These studies by Evans et al. [75], Zeltner and Anderson [38], and Lumsdon and Evans [36], illustrate the difficulty in removing all traces of carbonate from a solid–solution system, suggest that carbonate’s presence in ‘‘carbonate free’’ systems is a fairly common occurrence, and highlight the need for an approach to quantify the amount of residual carbonate present in a system. Therefore, given carbonate’s ubiquity and persistence in solid–solution systems, we contend that experimental data possessing carbonate contamination is commonly used in modeling studies, oftentimes without carbonate’s presence being taken into account. In this work, adsorption data taken from Barrow and Cox [5] and Bonnissel-Gissinger et al. [6] will be used to develop a CD-MUSIC model capable of predicting mercury uptake onto goethite. The presence of carbonate in both experimental systems will be considered in the model and quantified (i.e., ½CO2 3 TOT ) using an approach developed herein by the authors. To clarify, ½CO2 3 TOT denotes all carbonate present in the solid–solution system from both adsorbed (i.e., surface) and aqueous species. As stated previously, ion adsorption behavior in solid–solution systems is greatly influenced by the mineral surface’s crystal morphology [44–48]. One of the primary reasons for selecting CDMUSIC to use in this study is that the model takes into account the mineral surface’s heterogeneity. Therefore, in order to properly apply this adsorption model, the crystal face composition (CFC) of the mineral sample in question must first be determined. Unfortunately, neither of the mercury adsorption studies [5,6] employed in this work supplied a CFC for their particular goethite sample, BAR76 and BON15, respectively. Furthermore, determination of BAR76’s and BON15’s CFC through traditional means such as microscopic image analysis [40,49,50,53,54] or consideration of surface saturation data [47] was unattainable given the information provided by the authors. Although surface charging data for BAR76 and BON15 was presented in their respective studies, estimation of each sample’s CFC using the titration congruency method of Salazar-Camacho and Villalobos [48] could not be accomplished due to carbonate’s presence in the experimental systems of Barrow and Cox [5] and Bonnissel-Gissinger et al. [6]. Since it was not possible to determine BAR76’s and BON15’s CFC using any of the methods previously established (i.e. microscopic

image analysis, surface saturation data, or titration congruency), a new approach for estimating a sample’s CFC was needed. 2.3.2. New approach to estimating CFC Through examining goethite adsorption data found in the literature [40,44,46,49,51,53,76–78], Villalobos and co-workers [46,47,78] observed that, when normalized by surface area, goethite’s surface reactivity toward ions varies between samples. In particular, goethite samples with a SSA below 60 m2/g were found to have an appreciably higher reactivity [47,78]. To explain these findings, Villalobos and co-workers [46,47] hypothesized that as the mineral sample’s SSA decreases, the contribution from the (2 1 0) and (0 1 0) capping faces towards the sample’s CFC increases since these two crystal faces possess higher reactive site densities than the (1 0 1) and (0 0 1) faces (Table 2). This idea is supported by the findings of Weidler et al. [50] who in studying goethite’s surface morphology via atomic force microscopy (AFM), observed (2 1 0) step faces present on the (0 0 1) crystal face at such a high frequency that the (0 0 1) face was practically transformed into a (2 1 0) vicinal face. In other words, the (0 0 1) crystal face was blanketed by these (2 1 0) step faces to such a degree that the (0 0 1) face all but disappeared from goethite’s surface and was replaced by the (2 1 0) crystal face [47,50]. The goethite crystals examined by Weidler et al. [50] had a measured SSA of 7.9 m2/g. In contrast to these findings by Weidler et al. [50], goethite preparations with SSAs ranging from 94 to 105 m2/g presented in the literature are found to have surfaces comprised primarily of the (1 0 1) and (0 0 1) crystal faces (Table 4). From this evidence, Villalobos et al. [47] contended that as goethite crystals get larger and SSA decreases, the occurrence of these (2 1 0) step faces forming on the (0 0 1) crystal face increases, thereby elevating (2 1 0)’s contribution to the mineral surface. Furthermore, Villalobos et al. [47] postulated that the step faces observed by Weidler et al. [50] to be forming on the (0 0 1) face could be either the (2 1 0) or (0 1 0) crystal face. In fact, Villalobos et al. [47] used the (0 1 0) face in lieu of the (2 1 0) in characterizing goethite samples with SSAs of 70 and 50 m2/g. In their 2010 work, Salazar-Camacho and Villalobos [48] modeled arsenate adsorption onto goethite using previously published experimental data from several different studies. The authors

152

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161

Table 4 CFCs for goethite samples with SSAs between 94 and 105 m2/g. SSA (m2/g)

105a 96.4b 95c 95d 94e

CFC (%) (1 0 1)

(0 0 1)

(2 1 0)

90 90 90 70 70

– – – 30 30

10 10 10 – –

a–d

CFCs determined using microscopic image analysis. Hiemstra and Van Riemsdijk [49]. Rietra et al. [55]. c Venema et al. [25]. d Gaboriaud and Ehrhardt [40]. e CFC determined using chromate surface saturation data. Villalobos et al. [47]. a

b

estimated the CFC for a number of goethite samples, with SSAs ranging from 2–98 m2/g, via the titration congruency method, described in detail elsewhere (Salazar-Camacho and Villalobos [48]). None of the previously published adsorption studies that SalazarCamacho and Villalobos [48] used in their work provided surface saturation data. Furthermore, in nearly all of these studies, the sample goethite’s surface charging data was not presented. To circumvent this problem, Salazar-Camacho and Villalobos [48] utilized titration data from goethite samples possessing similar characteristics (i.e. SSA) and in similar environments (i.e. background electrolytes and ionic strength) as a surrogate for those samples that were missing surface charge data; thereby allowing for the CFC of each goethite sample to be ascertained via the titration congruency method. In a few cases, no reference goethite system with available titration data could be found to act as a reasonable substitute for the samples being studied. For these samples, their CFC was determined through modeling arsenate adsorption in a trial and error fashion. Using the CFCs determined by Salazar-Camacho and Villalobos [48], the proton reactive site density (NH), defined as the sum of reactive singly and triply coordinated surface sites per nm2 [23,25,46,47,52,54,57], was calculated for each goethite sample and then plotted versus SSA revealing a fairly linear relationship (Eq. (3), R2 = 0.8633) between the two variables and demonstrating that as SSA decreases, NH increases (Fig. 1). As seen in the data of Salazar-Camacho and Villalobos [48] and presented in Fig. 1 of this study, samples with a SSA 6 71 m2/g were found to possess a NH P 6.9 sites/nm2. Of the crystal faces considered to be prevalent on goethite’s surface [23,40,47,49–53], only the (2 1 0) and (0 1 0) crystal faces possess a NH greater than 6.9 sites/nm2 (Table 2). Given this information in conjunction with Fig. 1 and Eq. (3), it is evident that the contribution from the (2 1 0) and/or (0 1 0) crystal faces to the mineral’s CFC must increase as SSA decreases; a conclusion which further substantiates Villalobos’ and co-workers’ [46,47] hypothesis regarding increased reactivity for larger goethite crystals (lower SSA).

NH ¼ 0:028 ðSSAÞ þ 8:93

ð3Þ

The proton reactive site density for each goethite sample considered in this current modeling study was calculated using Eq. (3) along with the samples’ reported SSA (Table 1). Once a sample’s NH is determined, the CFC of the mineral can be calculated using the NH values of each crystal face, presented in Table 2, and assuming that the mineral surface is comprised of only two crystal faces. In this work the two crystal faces considered to be present on goethite’s surface were the (1 0 1) main face and the (0 1 0) capping face. The (1 0 1) crystal face was selected because of the amount of evidence supporting it as the dominant crystal face on goethite’s surface [50,51,79–81]. The (0 1 0) capping face was selected over the

(2 1 0) because only it possessed a high enough proton reactive site density to accommodate the NH calculated for BON15 (NH = 8.51 sites/nm2) using Eq. (3). The ability of Eq. (3) to predict a goethite sample’s NH and in turn, that sample’s CFC, was checked against previously published studies where the sample’s CFC had been determined via microscopic image analysis or surface saturation data. Through examination of electron micrographs, Venema et al. [54] estimated the CFC of a goethite sample, SSA 95 m2/g, to be composed of approximately 90% (1 0 1) main face and 10% (2 1 0) capping face. Applying Eq. (3) to this goethite sample results in a NH value of 6.28 sites/ nm2. Using this NH and assuming that only the (1 0 1) and (2 1 0) crystal faces are present on the mineral surface, a CFC of 85% (1 0 1) and 15% (2 1 0) results. In this case the (2 1 0) crystal face is considered rather than the (0 1 0) to stay consistent with Venema et al.’s [54] work. A 2009 study conducted by Villalobos et al. [47] estimated the CFC of a goethite sample with a SSA of 50 m2/ g using chromate surface saturation data. The resulting CFC determined by Villalobos et al. for this goethite sample yielded a NH value of 6.73 sites/nm2. In the same study, the CFC of a second goethite sample, also with SSA 50 m2/g, was established using chromate surface saturation data as well. For this case however, a different CFC was arrived at and the resulting NH was found to be 7.98 sites/nm2. When Eq. (3) is applied to goethite samples with a SSA of 50 m2/g, a NH of 7.53 sites/nm2 is calculated. The above examples are given to illustrate Eq. (3)’s usefulness in estimating the CFC of a goethite sample when information regarding the mineral surface is lacking; thereby reducing the number of adjustable parameters needed for the CD-MUSIC model to fit experimental data. Using Eq. (3) along with each sample’s SSA, and considering only the (1 0 1) and (0 1 0) crystal faces to be present on mineral’s surface, the CFCs for VIL70, BAR76, and BON15 were determined (Table 5). 2.3.3. Model description of the mineral surface’s heterogeneity Once the CFC for a goethite sample is known, the mineral surface’s heterogeneity can be quantitatively expressed in CD-MUSIC using one of two methods, the segregate approach or the composite approach (Table 6). In the segregate approach, each surface site type on the (1 0 1) and (0 1 0) crystal faces is input into CD-MUSIC as its own entity. There are a total of four site types considered using the segregate approach: (1) singly coordinated sites on the (1 0 1) face; (2) singly coordinated sites on the (0 1 0) face; (3) doubly coordinated sites on the (0 1 0) face; and (4) triply coordinated sites on the (1 0 1) face (Table 7). In the composite approach, all sites of a given type are considered to be identical to each other, regardless of which crystal face they are located on. Using the composite approach, there are three site types to take into account when modeling: (1) singly coordinated sites from the (1 0 1) and (0 1 0) crystal faces, (2) doubly coordinated sites from the (0 1 0) face, and (3) triply coordinated sites present on the (1 0 1) face (Table 6). Surface site types and densities determined using the segregate and composite approaches were employed separately in CDMUSIC to ascertain the impact that each surface characterization method had on the model’s predictive capabilities. Therefore, two unique adsorption models were developed in this study: (1) the

Table 5 Properties of the three goethite samples considered in this study. Goethite sample (sites/nm2)

SSA (m2/g)

NH

VIL70 BAR76 BON15

70 76 15

7.0 6.8 8.5

CFC (%) (1 0 1)

(0 1 0)

70 75 19

30 25 81

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161 Table 6 NS employed in CD-MUSIC using the segregate and composite approaches to describing the mineral’s surface. Goethite sample

Site type

BFeOH

VIL70

BFe2OH BFe3O BFeOH

BAR76

BFe2OH BFe3O BFeOH

BON15

BFe2OH BFe3O

Crystal face

NS (sites/nm2) Segregatea

Compositeb

(1 0 1) (0 1 0) (1 0 1) (0 1 0) (1 0 1) (0 1 0)

2.12 2.74 0.00 2.74 2.12 0.00

4.85

(1 0 1) (0 1 0) (1 0 1) (0 1 0) (1 0 1) (0 1 0)

2.29 2.23 0.00 2.23 2.29 0.00

(1 0 1) (0 1 0) (1 0 1) (0 1 0) (1 0 1) (0 1 0)

0.59 7.33 0.00 7.33 0.59 0.00

2.74 2.12

Dz0 ¼ n0 þ nH0  /ðn0 þ nH0 þ Rnref zref Þ

ð4Þ

2.23

Dz1 ¼ n1 þ nH1 þ /ðn0 þ nH0 þ Rnref zref Þ

ð5Þ

2.29

where n0 and n1 represent the portion of the adsorbate’s charge ascribed to the 0- and 1-planes, respectively. The nH0 and nH1 values represent the charge of additional protons positioned in the 0- and 1-planes, respectively. The symbol / is a proportionality factor (/  0.17). The term nref represents the number and zref the charge of the reference surface group(s) involved in the surface complexation reaction [60].

7.92 7.33 0.59

Table 7 The equilibrium constants of various Hg(II) aqueous species used in the CD-MUSIC model. log K

2+

+

+

Hg + H2O M HgOH + H Hg2+ + 2H2O M Hg(OH)2 + 2H+ + Hg2+ + 3H2O M HgðOHÞ 3 + 3H Hg2+ + Cl M HgCl+ Hg2+ + 2Cl M HgCl2  Hg2+ + 3Cl M HgCl3 2

Hg2+ + 4Cl M HgCl4 2+

Hg

+

+H +

CO23 

M HgHCOþ 3

determined using a similar method to the one discussed by Stachowicz et al. [82] and Hiemstra et al. [73]. In particular, CD values for Hg(II) surface species were determined by employing the Brown bond valence approach [58] in conjunction with Hg(II)Asurface oxygen bond lengths reported by Kim et al. [61,83] for Hg(II) adsorption onto goethite, in the presence and absence of chloride, measured via EXAFS spectroscopy. As described in detail by Hiemstra and Van Riemsdijk [60], for specifically adsorbed ions, the change of charge (Dz) on the 0- and 1-planes can be obtained using Eqs. (4) and (5), respectively:

4.52

a For each goethite sample, values under the column labeled ‘‘segregate’’ were calculated by multiplying the CFC percentage, presented in Table 5, for the crystal face being considered by that crystal face’s reactive site density (cf. Table 2) for the site type of interest (i.e. singly, doubly, or triply coordinated surface sites). b Values under the column labeled ‘‘composite’’ are determined for a given site type by summing the NS values presented in that site type’s row, under the ‘‘segregate’’ column for the (1 0 1) and (0 1 0) crystal faces.

Reactions

153

3.397 6.194 21.091 7.3 14 15 15.6 16.348

Hg2+ + CO23  M HgCO3(aq)

12.078

Hg2+ + 2CO23  M HgðCO3 Þ2 2 Hg2+ + H2O + Cl M HgOHCl + H+

15.578 4.25

segregate CD-MUSIC model; and (2) the composite CD-MUSIC model. For example consider the VIL70 goethite sample. In the ‘‘segregate’’ approach, the NS of singly coordinated sites from the (1 0 1) and (0 1 0) crystal faces is calculated by multiplying the CFC percentage for each face (70% and 30%, respectively), by the reactive NS of singly coordinated sites present on the (1 0 1) and (0 1 0) faces (3.03 and 9.1 sites/nm2, respectively). The resulting NS values of 2.12 and 2.74 sites/nm2, presented in the ‘‘segregate’’ column of Table 7, are summed together to give the NS of singly coordinated sites (4.85 sites/nm2) used in the composite approach and presented in the ‘‘composite’’ column. 2.3.4. Charge distribution (CD) values Charge distribution (CD) coefficients, Dz0 and Dz1, for innersphere surface complexes modeled in this work, were taken from the literature when available; however, for the adsorbates of concern in this study, the only CD values previously published were for carbonate surface species [67,68]. For mercury adsorption onto goethite, the CD coefficients for each complex considered were

3. Results and discussion 3.1. Dissolved mercury speciation Aqueous Hg(II) speciation was calculated as a function of pH and ionic strength using the equilibrium constants provided in MINEQL+ 4.5 [84,85] at 25 °C and listed in Table 7. In the absence of Cl and at low pH (<3), the hexaqua ion, HgðH2 OÞ2þ 6 , is the dominant mercury species. As pH increases, contributions from HgOH+ and Hg(OH)2 aqueous complexes increase. As Kim et al. [83] and others [3,8,10,11] have observed in systems free of complexing ligands and competing adsorbates, Hg(II) uptake onto goethite and other adsorbents peaks around pH 3 ± 1 and tapers off slightly as pH increases past this point. This low pH range at which the zenith of Hg(II) adsorption is attained coincides with Hg(OH)2 becoming the dominant mercury species in solution (Fig. 5b). As pH increases above the adsorption edge (pH > 4), the amount of Hg(II) adsorbed decreases marginally as a result of desorption to form the Hg(OH)2 aqueous complex [8] and an increase in weakly sorbing aqueous  species such as HgðOHÞ 3 [83]. In the presence of Cl , HgCl2 and HgOHCl can become dominant in the low to mid pH range thereby shifting the region of Hg(OH)2 dominance to higher pH values (Fig. 5d). 3.2. Surface protonation and ion pair formation The intrinsic constants for the protonation and ion pair formation reactions shown in Table 3 for the singly and triply coordinated surface sites were optimized with FITEQLC using the data from a carbonate-free goethite, VIL70 (13.8 g/L) in 0.1 and 0.015 M NaNO3 [32,33] (Fig. 2). The quality of the model fits to the data appears to be similar to those reported by Villalobos and Leckie [33] for the 2-pK TLM. In both cases, the models slightly over-predicted the surface charge at high pH and low ionic strength. 3.3. Carbonate species on goethite A number of studies investigating carbonate adsorption onto goethite using IR spectoscopy [33,64,86] have previously concluded that carbonate binds to the mineral’s surface as an innersphere monodentate complex. More recently however, Hiemstra et al.[67] re-examined these spectroscopic results and argued in

154

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161

Fig. 3. Mercury pH adsorption edge data for goethite sample BAR76 at four different total Hg(II) surface loadings in the absence of chloride. Symbols denote experimental data taken from Barrow and Cox [5] and lines denote CD-MUSIC model simulations using the (a) segregate and (b) composite approaches to describing the mineral’s surface (see text for more details). Reactions listed in Table 3 and [CO3]TOT values found in Table 8 were used in performing model simulations.

Fig. 4. Mercury pH adsorption edge data for goethite sample BAR76 at a Hg(II) surface loading of 0.03 lmol/m2 and five different total chloride surface loadings. Symbols denote experimental data taken from Barrow and Cox [5] and lines denote CD-MUSIC model simulations using the (a) segregate and (b) composite approaches to describing the mineral’s surface (see text for more details). Reactions listed in Table 3 and [CO3]TOT values found in Table 8 were used in performing model simulations.

favor of carbonate forming a bidentate inner-sphere complex with goethite’s surface based on the magnitude of carbonate’s Dv3 band splitting, the charge distribution values determined for carbonate via surface complexation modeling, and comparison of proton coadsorption data between carbonate and other oxyanions such as selenite. Further evidence in support of the bidentate structure proposed by Hiemstra et al. [67] was provided by Bargar et al. [41], who found carbonate to form a binuclear bidentate complex on hematite through the use of attenuated total reflectance Fourier transform infrared (ATR–FTIR) spectroscopy and quantum chemical calculations. In addition to the bidentate inner-sphere surface species, Bargar et al. [41] also observed a CAO stretching band around 1460 cm1 in D2O that was attributed to the formation of an outer-sphere carbonate complex on hematite and confirmed with molecular orbital calculations using density functional theory (MO/DFT). Wijnja and Schulthess [86] observed a similar CAO stretching band (1462 cm1 in D2O) when investigating carbonate adsorption onto goethite using ATR-FTIR spectroscopy. Given this similarity between the CAO stretching bands, Rahnemaie et al. [68] proposes that the same outer-sphere complex Bargar et al. [41] found to form on hematite also forms on goethite. The carbonate surface species employed by Rahnemaie et al. [68], which are supported by spectroscopic evidence and MO/DFT calculations, were considered in developing the CD-MUSIC model used in our work. To account for the presence of carbonate in the solid–solution systems of Barrow and Cox [5] and Bonnissel-Gissinger et al. [6], carbonate adsorption was incorporated into our CD-MUSIC model and calibrated using experimental data provided in Villalobos and Leckie [33]. Using CD-MUSIC to model this data in conjunction

with the carbonate surface complexes employed by Rahnemaie et al. [68] allowed for identification of the relevant carbonate surface species and their intrinsic formation constants. The best fits of Villalobos’ and Leckie’s [33] experimental data were obtained when only the inner-sphere bidentate complex was utilized in our model (Table 3); a result that is in agreement with Hiemstra et al. [67] who successfully modeled the same experimental data using only the inner-sphere bidentate complex as well. When employing the segregate approach to characterizing goethite’s surface (Table 6), the CD-MUSIC model worked best when the carbonate surface complex only formed on the (010) crystal face (Table 3). At higher surface loadings, adsorption of carbonate onto the (1 0 1) crystal face is expected to occur; however, for the experimental data modeled from Villalobos and Leckie [33], the formation of an inner-sphere bidentate complex on the (1 0 1) crystal face was not necessary since there were sufficient sites present on the (0 1 0) face to accommodate all carbonate adsorbed and these sites possess a higher reactivity than those on the (1 0 1) face [87,88]. It is important to note that carbonate’s inner-sphere bidentate complex is thought to be a double-corner bidentate complex, binding to two singly coordinated surface sites present on the corners of adjacent octahedra [41,67,68]. While this configuration is possible on the (1 0 1), (0 0 1), and (2 1 0) crystal faces, it is not attainable on the (0 1 0) face because singly coordinated sites are present as pairs on the edges of the same octahedra rather than corners on adjacent octahedra [52]. As illustrated by SalazarCamacho and Villalobos [48], on the (0 1 0) face, the corners of adjacent octahedra are comprised of one singly coordinated site and one doubly coordinated site. Hence, in order to model a double corner bidentate complex on the (0 1 0) crystal face, carbonate

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161

155

Fig. 5. (a) Surface speciation of mercury on BAR76 for a pH adsorption edge experiment in the absence of chloride (Hg(II) surface loading of 0.03 lmol/m2). The symbols represent experimental data points for total Hg(II) adsorbed. The solid line is taken from the model simulation and denotes the total Hg(II) adsorbed. In the absence of chloride, all mercury adsorbs to goethite as („FeO)AHgA(„Fe2O). The segregate method of characterizing the mineral’s surface was used in CD-MUSIC for this model simulation. (b) Aqueous speciation of Hg(II) in solution during the pH adsorption edge experiment presented in plot a. (c) Surface speciation of Hg(II) on BAR76 in a model simulation of a pH adsorption edge in the presence of chloride (Hg(II) and Cl surface loadings of 0.03 and 1.64 lmol/m2). The symbols represent experimental data points for total Hg(II) adsorbed. The lines are taken from the model simulation with the solid line denoting the total Hg(II) adsorbed and the dashed lines representing the surface species used in the CD-MUSIC model. The segregate method of characterizing the mineral’s surface was used in CD-MUSIC for this model simulation. (d) Aqueous speciation of Hg(II) in solution during the pH adsorption edge experiment presented in plot c.

must bind to one singly and one doubly coordinated site. Therefore, in the segregate CD-MUSIC model employed in this work, carbonate was modeled as complexing with one singly and one doubly coordinated site on the (0 1 0) crystal face of goethite; thereby allowing for the model to agree with spectroscopy and molecular modeling simulations. With the relevant carbonate surface species known and the intrinsic log K values for surface protonation, ion pair formation, and carbonate surface complexation reactions determined from modeling the data provided in Villalobos and Leckie [33], the CDMUSIC model could now predict the carbonate contaminated potentiometric titration data of Barrow and Cox [5]. Since the extent of carbonate contamination was not known, the value of ½CO2 3 TOT present in the solid–solution system was optimized using FITEQLC to achieve the best model fits of the titration data. In performing this modeling procedure it was discovered that the optimal ½CO2 3 TOT value varied with the ionic strength of the system and that one ½CO2 3 TOT value could not satisfactorily describe all three potentiometric titration data sets (ionic strengths of 0.001, 0.01, and 0.1 M NaNO3) presented in Barrow and Cox [5]. Given that the Hg(II) adsorption experiments of Barrow and Cox [5] were all conducted at an ionic strength of 0.1 M NaNO3, the optimized ½CO2 3 TOT value obtained for the titration data set possessing that same ionic strength was utilized in all CD-MUSIC simulations of Hg(II) adsorption onto BAR76 (Table 3). For the BAR76 solid–solution system at 0.1 M NaNO3, the opti4 mized ½CO2 M and 3 TOT values were determined to be 2.25  10 4 2.38  10 M for the segregate and composite CD-MUSIC models,

respectively. Model calculations using the conditions described by Barrow and Cox [5] (i.e., [Hg(II)]TOT = 0–50 lM, [Cl]TOT = 0–5000 lM, [goethite] = 4.02–4.51 g/L, ionic strength of 0.1 M NaNO3) reveal that both of these optimized ½CO2 3 TOT values are in agreement with the experimental system being in equilibrium with the atmosphere (PCO2 = 0.38 matm) at pH values around 7. In other words, the optimized ½CO2 3 TOT values for the BAR76 solid–solution system, obtained using the procedure outlined above, are equivalent to the ½CO2 3 TOT concentrations predicted for the experimental system when constrained by equilibrium with the atmosphere. It should be noted that to determine the value of ½CO2 3 TOT present in the BAR76 solid–solution system under atmospheric CO2 conditions, concentrations of the aqueous species CO2 3 in equilibrium with atmospheric CO2(g) were first determined at varying pH values using the chemical equilibrium software Visual MINTEQ 3.0 [69]. These concentrations of the CO2 3 aqueous species and their corresponding pH values were then input into FITEQL as serial data. Subsequently, the segregate and composite CD-MUSIC models were run in FITEQL to determine the total amount of carbonate (i.e., ½CO2 3 TOT ) predicted to be present in the system, when open to the atmosphere, at each pH value considered. It is important to clarify that in Bonnissel-Gissinger et al.’s [6] study, no potentiometric titration data was presented; rather, the reported pHPZC of the BON15 goethite sample was based on a previous study conducted by Mueller and Sigg [34] who utilized the same goethite preparation, Bayferrox 910 (standard 86), from Bayer. With no potentiometric titration data available for the goethite sample actually used in Bonnissel-Gissinger et al.’s [6] work,

156

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161

Table 8 Summary of CD-MUSIC model fits to Hg(II) adsorption data for BAR76 and BON15. Goethite sample

BAR76

Electrolyte

NaNO3

NaNO3 and NaCl

BON15

NaNO3 and NaCl

Surface loading

Model fit 

Hg(II)

Cl

Segregate

Composite

0.01 0.03 0.07 0.15 0.03 0.03 0.03 0.03

– – – – 0.016 0.164 1.64 16.4

19 3.4 6.3 5.0 3.7 15 12 2.9

17 2.8 4.2 5.6 5.6 12 11 2.7

0.08 0.08 0.08 0.08

0.13 0.13 667 667

5.9 10 0.5 0.7

3.4 3.4 1.3 1.1

‘‘Segregate’’ and ‘‘Composite’’ headings denote which surface characterization method was employed in the CD-MUSIC model (see text for more details). Surface loadings are 4 in units of lmol/m2. Model fits were calculated with the WSOS/DF parameter. Using the segregate method for BAR76 and BON15, ½CO2 M and 3 TOT = 2.25  10 4 5  3.42  103 M, respectively. Using the composite method for BAR76 and BON15, ½CO2  = 2.38  10 M and 5.43  10 M, respectively. The symbol denotes model 3 TOT simulations performed assuming no carbonate was present in the system. All experiments were conducted in a background electrolyte solution of 0.1 M ionic strength.

it is difficult to ascertain the existence and extent of carbonate contamination present in the solid–solution systems they studied based solely on another study’s [34] titration data. Fortunately, Bonnissel-Gissinger et al. [6] utilized X-ray photoelectron spectroscopy (XPS) to determine the cleanliness of their mineral sample’s surface and upon analyzing BON15’s XPS spectrum, observed carbon to be present on the mineral’s surface; thereby providing direct evidence of carbonate contamination on BON15. It should be noted that Bonnissel-Gissinger et al. [6] stipulate that prior to experimentation, their goethite preparation was washed with degassed Milli-Q water and dried in a dessicator to prevent carbonate formation on the mineral surface. Given all this information, specifically the pHPZC referenced by the authors, the XPS results, and the statement by the authors regarding the steps taken to ensure experiments were conducted in a carbon free environment, it is unclear whether or not carbonate was present in the system. To be thorough in our investigation, both scenarios (i.e. with and without carbonate present in the system) were considered separately when modeling the Hg(II) adsorption data presented in Bonnissel-Gissinger et al. [6]. For the scenario where carbonate is present, ½CO2 3 TOT was optimized using FITEQLC to achieve the best model fits of the Hg(II) adsorption data (Table 8).

3.4. Hg(II) adsorption on goethite Adsorption modeling for Hg(II) onto BAR76 incorporated competition with carbonate species since, as stated previously, it was reasoned that the experimental systems of Barrow and Cox [5] were contaminated with carbonate. The CD-MUSIC model was initially calibrated using data collected for Hg(II) adsorption in the absence of chloride. A subsequent model calibration was performed in the presence of Cl to determine the intrinsic equilibrium constant for the HgACl ternary surface complex. All parameters and constants optimized from the previous simulations concerning potentiometric titration and carbonate adsorption data were integrated into the model as fixed parameters including aqueous complexation equilibrium constants and the Stern layer capacitances. A total of eight Hg(II) pH adsorption edges, four with and four without chloride present in solution, were successfully predicted on BAR76 using the 1-pK CD-MUSIC model (Figs. 3 and 4). Using EXAFS analysis in conjunction with molecular modeling simulations to investigate Hg(II) adsorption onto goethite, both Kim et al. [83] and Collins et al. [89] found strong evidence suggesting that Hg(II) binds to the mineral surface as an inner-sphere binuclear bidentate corner sharing complex via two deprotonated

singly coordinated surface oxygens. Utilizing bond valence analysis as a complement to their EXAFS and molecular modeling work, Kim et al. [83] found that Hg(II) bound to either singly protonated or deprotonated singly coordinated surface oxygens („FeOH and „FeO, respectively) resulted in a saturated oxygen coordinative state signifying a stable surface species. In other words, bond valence calculations indicate that the two singly coordinated surface sites that bind with Hg(II) to form the binuclear bidentate surface complex can be singly protonated or deprotonated. This result contradicts molecular modeling simulations that found only deprotonated singly coordinated sites complexing with Hg(II) are capable of achieving the HgAO interatomic distance observed using EXAFS spectroscopy [83,89]. Given that bond valence analysis has been employed in a number of studies to help guide selection of surface species based on EXAFS data [62–64], the bond valence findings of Kim et al. [83] should not be immediately dismissed despite the disagreement with molecular modeling results; therefore, three different Hg(II) bidentate surface complexes were considered in developing the composite CD-MUSIC model: („FeOH)2AHg, („FeO)2HAHg, and („FeO)2AHg. Upon modeling the experimental data of Barrow and Cox [5] with CD-MUSIC, it was found that both the („FeO)2HAHg and („FeO)2AHg complexes were needed to describe Hg(II)’s adsorption behavior onto goethite (cf. Table 3). In the case of the segregate CD-MUSIC model, Hg(II) was only considered to complex with the (0 1 0) crystal face. The reason for this is that surface sites present on the (0 1 0) face are thought to be more reactive than those on the (1 0 1) face [87,88], and, for the experimental conditions tested by Barrow and Cox [5] and Bonnissel-Gissinger et al. [6], all adsorbed Hg(II) can be accommodated on the (0 1 0) crystal face. Therefore, it stands to reason that, whenever possible, all Hg(II) would preferentially adsorb to the (0 1 0) face. To be sure, Hg(II) is expected to complex with the (1 0 1) face at higher surface loadings; however, under the conditions considered in this work, that is not believed to occur. In the segregate CD-MUSIC model, Hg(II) complexes with the (0 1 0) face as a double corner bidentate complex to one singly and one doubly coordinated surface site (i.e., („FeO)AHgA(„Fe2O)). This double corner bidentate complex was chosen because it provided the best agreement with the spectroscopic observations, molecular modeling simulations, and bond valence analysis of Kim et al. [83]. In the presence of chloride, Barrow and Cox [5] and Gunneriusson and Sjoeberg [8] both observed enhanced Hg(II) uptake onto goethite that could not be explained without the addition of a HgACl ternary surface complex. To confirm the formation of this proposed surface species, Kim et al. [61] conducted a series of Hg(II) uptake measurements onto goethite with varying levels of

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161

chloride present in solution and examined the resulting sorption products using EXAFS spectroscopy. Through interpretation of their spectroscopic data and confirmation via molecular modeling simulations, the authors concluded that a Type A (surface–metal– ligand) HgACl ternary surface complex forms on goethite in a monodentate fashion; in agreement with the findings of Bargar et al. [4]. The HgACl surface species complexed to a deprotonated singly coordinated surface oxygen provided the best agreement between EXAFS data and molecular modeling results [61], and was therefore employed in the current modeling study (Table 3). CD-MUSIC was able to provide satisfactory model fits (0.1 < WSOS/DF < 20) of all mercury pH adsorption edge experi-

157

ments conducted on BAR76 using either surface characterization method, segregate or composite, employed in this study (Figs. 3 and 4). Whether in the absence or presence of chloride, Hg(II) adsorption was found to peak around the same pH that Hg(OH)2 became the dominant mercury species in solution (Fig. 5). In the presence of chloride, the adsorption edge shifted to a higher pH because of the dominance of the non-sorbing aqueous species HgCl2 in the lower pH range. From the model results seen in Fig. 5, the HgOHCl aqueous complex appears to be the sorbing species responsible for the formation of the „FeOHgCl surface complex given how the concentrations of the aqueous and surface species coincide well

Fig. 6. (a and b) Effect of [CO3]TOT on predicted mercury adsorption to BAR76 in the absence of chloride using the CD-MUSIC model with the (a) segregate and (b) composite approaches, respectively, for describing the mineral’s surface (see text for more details). Simulations were conducted using a Hg(II) surface loading of 0.07 lmol/m2. Using the chemical equilibrium software Visual MINTEQ 3.0 [69], no precipitates were predicted to form under the conditions tested in these simulations. (c) and (d) Percent of 2 Hg(II)TOT present in the system as an aqueous carbonate complex (i.e., HgHCOþ 3 , HgCO3, and HgðCO3 Þ2 ) for CD-MUSIC simulations presented in plot a using the segregate approach and plot b using the composite approach, respectively. (e) and (f) Percent of Hg(II) singly coordinated surface sites („FeOH) occupied by carbonate for CD-MUSIC simulations presented in plot a using the segregate approach and plot b using the composite approach, respectively. Since carbonate only binds to the (0 1 0) crystal face in the segregate CD-MUSIC model, only „FeOH sites on the (0 1 0) crystal face were considered in calculating the percent „FeOH sites occupied in plot e. The labels and corresponding arrows used in all plots are to denote the carbonate conditions used in each model simulation.

158

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161

with each other. As pH increases in systems containing chloride, Hg(OH)2 eventually becomes the dominant mercury species in solution; the rise of this aqueous complex (Fig. 5d) corresponds fittingly with the rise in concentration of the („FeO)AHgA(„Fe2O) surface complex (Fig. 5c). Using the optimized ½CO2 3 TOT values determined for the BAR76 4 solid–solution system (½CO2 M and 2.38  104 M 3 TOT = 2.25  10 for the segregate and composite CD-MUSIC models, respectively), both models predicted carbonate to have minimal influence on Hg(II)’s adsorption to goethite (Fig. 6). To ascertain what impact, if any, carbonate has on mercury uptake in the absence of chloride, model simulations employing higher ½CO2 3 TOT values were conducted for the BAR76 system using a Hg(II) surface loading of 0.07 lmol/m2. At elevated ½CO2 3 TOT concentrations, both the segregate and composite CD-MUSIC models predict the mercury adsorption edge to not only elongate, but also become increasingly sinusoidal in appearance as ½CO2 3 TOT rises (Fig. 6a and b). In both models, Hg(II) and carbonate compete for the same reactive surface sites (Table 3); therefore, as ½CO2 3 TOT increases, more carbonate will bind to the goethite sample thereby reducing the number of surface sites available for Hg(II) adsorption (Fig. 6e and f). Comparison of Fig. 6a and e, and Fig. 6b with f, reveal that the pH around which the Hg(II) adsorption edge experiences a minimum (i.e., the trough of the wave) corresponds well with the pH at which maximum carbonate adsorption occurs. In addition to competition with carbonate for surface sites, Hg(II)’s uptake onto BAR76 may also be affected by its aqueous speciation with carbonate (Fig. 6c and d). As the ½CO2 3 TOT value rises, Hg(II) becomes increasingly complexed 2 with carbonate in solution (i.e., HgHCOþ 3 , HgCO3, and HgðCO3 Þ2 aqueous complexes) (Fig. 6c and d). As stated previously, evidence suggest that mercury’s bidentate complex forms on goethite’s surface primarily by means of the Hg(OH)2 aqueous complex (Fig. 5a and b); therefore, it is reasoned that these mercury carbonate

aqueous complexes are effectively trapping Hg(II) in solution and preventing its adsorption to goethite. 3.5. Model verification Mercury adsorption data taken from Bonnissel-Gissinger et al. [6] was used to verify the predictive capability of the CD-MUSIC model developed in this study. Model simulations were conducted using the reactions and affinity constants reported in Table 3. For both segregate and composite surface characterizations of BON15, the CD-MUSIC model was able to satisfactorily simulate Hg(II) uptake onto the goethite sample for the experimental conditions tested by Bonnissel-Gissinger et al. [6] (Fig. 7). In contrast to the modeling results of BAR76, the CD-MUSIC model utilizing the composite method produced the best model fits for Hg(II) adsorption onto BON15. As discussed previously, two scenarios, one with and one without carbonate present in the BON15 solid–solution system, were considered when modeling the Hg(II) adsorption data of Bonnissel-Gissinger et al. [6]. Examination of Fig. 7 and Table 8 reveals that the difference in model fit between the two scenarios tested was minimal; suggesting that, under the conditions tested 3 (½CO2 M and 5.43  105 M for the segregate 3 TOT = 3.42  10 and composite CD-MUSIC models, respectively), carbonate’s presence in the system has little to no impact on Hg(II)’s adsorption behavior onto goethite. Similar to the analysis performed on the BAR76 system, model simulations were carried out for the BON15 solid–solution system at elevated ½CO2 3 TOT values to ascertain what impact, if any, carbonate has on mercury adsorption in the presence of chloride (Fig. 8a and b). Simulations were performed using Hg(II) and chloride surface loadings of 0.08 and 0.13 lmol/ m2, respectively. As ½CO2 3 TOT increases, both the segregate and composite CDMUSIC models predict the mercury adsorption edge to elongate

Fig. 7. Mercury pH adsorption edge data for goethite sample BON15 at a Hg(II) surface loading of 0.08 lmol/m2 and two different total chloride surface loadings. Symbols denote experimental data taken from Bonnissel-Gissinger et al. [6] and lines denote CD-MUSIC model simulations using the [(a) and (c)] segregate and [(b and d)] composite approaches to describing the mineral’s surface (see text for more details). Reactions listed in Table 3 and [CO3]TOT values of 3.42  103 M for plot a and 5.43  105 M for plot b, found in Table 8, were used in performing the model simulations. Model simulations for plots c and d were conducted assuming no carbonate was present in the system.

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161

and shift to higher pH values; however, one discernible difference between the two models’ simulations is the degree of elongation. In particular, whereas the adsorption edge predicted by the composite CD-MUSIC model becomes increasingly elongated as ½CO2 3 TOT rises, the segregate version of the model predicts mercury’s adsorption edge to shift to higher pH values while maintaining a similar shape when compared with the adsorption edge simulated under experimental conditions (Fig. 7a and Fig. 8a, seg3 regate CD-MUSIC model, ½CO2 M; Fig. 7b and 3 TOT = 3.42  10 5 Fig. 8b, composite CD-MUSIC model, ½CO2  M). 3 TOT = 5.43  10 For both models, the reason mercury’s adsorption edge is predicted to shift towards higher pH values and elongate with higher ½CO2 3 TOT values is due to increased competition between Hg(II)

159

and carbonate for the same surface sites (Table 3). For the segregate and composite CD-MUSIC models, carbonate competes with mercury surface species for reactive sites present on BON15’s surface, resulting in decreased amounts of each Hg(II) surface complex formed (Fig. 8c–f). In contrast to the model results found for the BAR76 system, mercury adsorption to BON15 is not heavily influenced by complexation to carbonate for either the segregate or composite CDMUSIC models due to the presence of chloride in the system which acts as a much stronger ligand to Hg(II) than carbonate. At one of 3 the higher ½CO2 M) for the BON15 sys3 TOT values tested (5  10 tem, the maximum percentage of Hg(II)TOT complexed with carbonate in solution (max HgðIIÞTOT;CO3 ) was found to be 1% for the

Fig. 8. (a and b) Effect of [CO3]TOT on predicted mercury adsorption to BON15 in the presence of chloride using the CD-MUSIC model with the (a) segregate and (b) composite approaches to describing the mineral’s surface (see text for more details). Simulations were conducted using Hg(II) and chloride surface loadings of 0.08 and 0.13 lmol/m2, respectively. Using the chemical equilibrium software Visual MINTEQ 3.0 [69], no precipitates were predicted to form under the conditions tested in these simulations. (c and d) Predicted concentrations of the HgCl ternary surface complex („FeOAHgCl), in units of lmol/m2, for the CD-MUSIC model simulations presented in plots a and b, respectively. (e and f) Predicted concentrations of („FeO)AHgA(„Fe2O) for the segregate CD-MUSIC model and („FeO)2HAHg and („FeO)2AHg for the composite CDMUSIC model, in units of lmol/m2, for the model simulations presented in plots a and b, respectively. The labels and corresponding arrows used in all plots are to denote the carbonate conditions used in each model simulation.

160

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161

segregate model and 2% for the composite model. Using that same 3 ½CO2 M) for the BAR76 system, in the absence 3 TOT value (5  10 of chloride, the max Hg(II)TOT,CO3 value was predicted to be approximately 29% in the segregate and composite models. For the BON15 system, optimized ½CO2 3 TOT concentrations were calculated using the mineral sample’s mercury adsorption data. Hg(II) uptake onto BON15 was predicted using the segregate and composite CD-MUSIC models along with the surface complexes and equilibrium constants previously determined from fitting the adsorption data of VIL70 and BAR76 (Table 3). Since the amount of carbonate present in the BON15 solid–solution system was unknown, the value of ½CO2 3 TOT was optimized using FITEQLC to achieve the best model fits of the mercury adsorption data. The difference between the two CD-MUSIC models optimized ½CO2 3 TOT 3 concentrations for the BON15 system (½CO2 M 3 TOT = 3.42  10 5 and 5.43  10 M for the segregate and composite CD-MUSIC models, respectively), stems from the fact that the segregate CDMUSIC model possesses a larger log K value than the composite CD-MUSIC model for the „FeOAHgCl ternary surface complex (Table 3). As a result of this larger log K value, the segregate CD-MUSIC model predicts more mercury to bind to BON15 (chloride surface loading, CCl, of 0.13 lmol/m2) at the lower end of the pH adsorption edge than the composite CD-MUSIC model (Fig. 7 and Fig. 8c and d). In fact, in the absence of carbonate, the segregate CD-MUSIC model over predicts mercury adsorption at pH values below 5; while, within that same pH range, the composite CD-MUSIC model’s prediction of Hg(II) uptake is closer in line with the observed experimental values (Fig. 7c and d). As previously discussed, when carbonate is present in the segregate CD-MUSIC model, it competes with Hg(II) for reactive surface sites present on the (0 1 0) crystal face. For the BON15 system (CCl = 0.13 lmol/m2), increasing ½CO2 3 TOT concentrations in the segregate CD-MUSIC model will reduce the amount of mercury predicted to adsorb (Fig. 8a). In a similar way, carbonate’s presence in the composite CD-MUSIC model results in a predicted decrease in mercury adsorption (Fig. 8b). Given carbonate’s influence on mercury’s predicted adsorption behavior for both CD-MUSIC models considered, and in light of the model predictions for Hg(II) adsorption onto BON15 (CCl = 0.13 lmol/m2) in the absence of carbonate (Fig. 7c and d); the optimized ½CO2 3 TOT values can be better understood. Recall that in a carbonate free simulation of BON15 (CCl = 0.13 lmol/ m2), the segregate CD-MUSIC model was found to over predict mercury adsorption below pH 4.5. Introducing carbonate into the segregate CD-MUSIC model at a concentration of ½CO2 3 TOT = 3.42  103 M, causes the predicted mercury adsorption edge to shift downward, thereby allowing for the model to better align with the experimental data, specifically at the lower end of the pH adsorption edge. Using the composite CD-MUSIC model, mercury adsorption is slightly over predicted at pH values below 4.5 and under predicted above pH 4.5 for the carbonate free simulation of the BON15 system (CCl = 0.13 lmol/m2; Fig. 7d). Since carbonate’s presence in the system shifts the mercury adsorption edge downward (Fig. 8b); the optimized ½CO2 3 TOT value for the composite model needs to be small, relative to the segregate model’s optimized ½CO2 3 TOT concentration, to avoid diminishing the model’s accuracy above pH 4.5 by causing further under predictions of Hg(II) adsorption onto BON15. For this reason, the optimized ½CO2 3 TOT value for the composite CD-MUSIC model was found to be 5.43  105 M, resulting in no discernible difference in the model’s predicted mercury adsorption edge for the carbonate free simulation and the optimized ½CO2 3 TOT simulation (Fig. 8b). It should be noted that the optimized ½CO2 3 TOT values calculated for the segregate and composite CD-MUSIC models (3.42  103 M and 5.43  105 M, respectively) are both in agreement with the BON15 solid–solution system being constrained by equilibrium with the atmosphere.

4. Conclusions The adsorption behavior of Hg(II) on two different goethite samples in the presence of chloride and carbonate was accurately predicted using the CD-MUSIC model. Not only was the model’s predicted removal of Hg(II) consistent with the experimental data tested, but the speciation of mercury in the aqueous phase helped to explain the impacts of pH, chloride, and carbonate on Hg(II) uptake. Model simulations conducted at elevated ½CO2 3 TOT values reveal that carbonate does indeed influence mercury adsorption onto goethite. In particular, competition between carbonate and Hg(II) for the same reactive sites on the mineral surface results in mercury’s adsorption edge elongating and shifting to higher pH values as the concentration of ½CO2 3 TOT increases. The CD-MUSIC model developed in this work is entirely selfconsistent. In all model simulations performed, the same affinity constants for surface protonation, ion pair formation, carbonate adsorption, and mercury surface complexation reactions were utilized. Spectroscopic evidence and molecular modeling studies found in the literature were used to guide the selection of surface species incorporated into the adsorption model. The extent of carbonate contamination in the solid–solution systems studied was determined using a novel approach developed in this work. The ½CO2 3 TOT values predicted using this methodology provided good agreement with the experimental systems being constrained by equilibrium with the atmosphere. The CFC of each goethite sample studied was estimated using a new approach based off the work of Salazar-Camacho and Villalobos [48] that only requires knowledge of the mineral sample’s SSA. In successfully simulating Hg(II) adsorption onto BAR76 and BON15, the CD-MUSIC model developed here has proven capable of describing Hg(II) adsorption onto goethite samples with varying surface reactivity in relatively complex, multi-solute systems. Furthermore, the success of the CDMUSIC model demonstrates the applicability of the CFC estimation technique proposed in this study and its utility when there is limited information available concerning the adsorbent. With the composite CD-MUSIC model’s success in satisfactorily fitting all experimental data tested in this study, it is tempting to suggest that the composite surface characterization method should be employed more often in CD-MUSIC models; given that it is more simplistic than the segregate approach and easier to model. However, it is important to recall that while some of the more simplistic surface complexation models (SCMs) such as the diffuse layer model (DLM) have proven capable of correctly predicting metal ion adsorption behavior in certain instances, their deficiencies have also been demonstrated for increasingly complex systems [90]. Of the two surface characterization methods considered in this study, the segregate approach provides a more detailed description of the mineral surface’s heterogeneity and permits greater specificity within the CD-MUSIC model regarding each surface complex’s bonding environment; thereby allowing for the modeled surface species to be in closer agreement with molecular scale analyses (i.e., spectroscopy, molecular modeling). As stated previously, ion adsorption behavior in solid–solution systems is greatly influenced by the mineral surface’s crystal morphology [44–48]; hence, it would be expected that the more accurately a model can describe a mineral surface, the better its predictive capability becomes. Therefore, while both CD-MUSIC models were successful in simulating Hg(II) adsorption onto BAR76 and BON15, and are suitable for use within the experimental conditions tested here; we speculate that the segregate version of the CD-MUSIC model is better suited for predicting Hg(II) adsorption in more complex systems with variable environmental conditions.

J.E. Mangold et al. / Journal of Colloid and Interface Science 418 (2014) 147–161

Acknowledgments The authors thank the U.S. Department of Energy, Office of Basic Energy Sciences, and Division of Chemical Sciences, Geosciences, and Biosciences both at Sandia National Laboratories and through Grant No. DE-FG02-04ER15495 at UT-Austin. This work is also supported by the royalty income from TerraTherm which licensed the ISTD technology after Shell’s donation of the patents to The University of Texas at Austin. The authors would also like to thank Mario Villalobos for his help in explaining the titration congruency method. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43]

U.S. EPA; 2006. P. Schindler, Adv. Chem. Ser. No. 67 (1967) 196. D. Sarkar, M.E. Essington, K.C. Misra, Soil Sci. Soc. Am. J. 63 (1999) 1626. J.R. Bargar, P. Persson, G.E. Brown Jr., J. Phys. IV 7 (1997) 825. N.J. Barrow, V.C. Cox, J. Soil Sci. 43 (1992) 295. P. Bonnissel-Gissinger, M. Alnot, J.-P. Lickes, J.-J. Ehrhardt, P. Behra, J. Colloid Interface Sci. 215 (1999) 313. J.A. Davis, R.O. James, J.O. Leckie, J. Colloid Interface Sci. 63 (1978) 480. L. Gunneriusson, S. Sjoeberg, J. Colloid Interface Sci. 156 (1993) 121. D.G. Kinniburgh, M.L. Jackson, Soil Sci. Soc. Am. J. 42 (1978) 45. R.A. Lockwood, K.Y. Chen, Environ. Lett. 6 (1974) 151. M.G. MacNaughton, R.O. James, J. Colloid Interface Sci. 47 (1974) 431. D. Sarkar, M.E. Essington, K.C. Misra, Soil Sci. Soc. Am. J. 64 (2000) 1968. C. Tiffreau, J. Luetzenkirchen, P. Behra, J. Colloid Interface Sci. 172 (1995) 82. T. Viraraghavan, A. Kapoor, Appl. Clay Sci. 9 (1994) 31. R. Weerasooriya, D. Aluthpatabendi, H.J. Tobschall, Colloids Surf., A 189 (2001) 131. R. Weerasooriya, B. Dharmasena, D. Aluthpatabendi, Colloids Surf., A 170 (2000) 65. R. Weerasooriya, H.J. Tobschall, A. Bandara, Chemosphere 69 (2007) 1525. R. Weerasooriya, H.J. Tobschall, H.K.D.K. Wijesekara, E.K.I.A.U.K. Arachchige, K.A.S. Pathirathne, Chemosphere 51 (2003) 1001. Y. Yin, H.E. Allen, Y. Li, C.P. Huang, P.F. Sanders, J. Environ. Qual. 25 (1996) 837. K.F. Hayes, G. Redden, W. Ela, J.O. Leckie, J. Colloid Interface Sci. 142 (1991) 448. I. Rojo, F. Seco, M. Rovira, J. Gimenez, G. Cervantes, V. Marti, J. Nucl. Mater. 385 (2009) 474. D.A. Sverjensky, K. Fukushi, Environ. Sci. Technol. 40 (2006) 263. T. Hiemstra, P. Venema, W.H. Van Riemsdijk, J. Colloid Interface Sci. 184 (1996) 680. J. Lützenkirchen, Environ. Sci. Technol. 32 (1998) 3149. P. Venema, T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interface Sci. 183 (1996) 515. P. Venema, T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interf Sci. 192 (1997) 94. H. Farrah, W.F. Pickering, Air, Soil Pollut. 9 (1978) 23. E.A. Forbes, A.M. Posner, J.P. Quirk, J. Colloid Interface Sci. 49 (1974) 403. G.E. Jean, G.M. Bancroft, Geochim. Cosmochim. Acta 50 (1986) 1455. D.W. Newton, R. Ellis Jr., G.M. Paulsen, J. Environ. Qual. 5 (1976) 251. M.G. MacNaughton, Stanford University, Palo Alto, CA, 1973. M. Villalobos, J.O. Leckie, Geochim. Cosmochim. Acta 64 (2000) 3787. M. Villalobos, J.O. Leckie, J. Colloid Interface Sci. 235 (2001) 15. B. Mueller, L. Sigg, J. Colloid Interface Sci. 148 (1992) 517. G. Sposito, Environ. Sci. Technol. 32 (1998) 2815. D.G. Lumsdon, L.J. Evans, J. Colloid Interface Sci. 164 (1994) 119. G.A. van, A.P. Robertson, J.O. Leckie, Geochim. Cosmochim. Acta 58 (1994) 2073. W.A. Zeltner, M.A. Anderson, Langmuir 4 (1988) 469. J.D. Filius, T. Hiemstra, R.W.H. Van, J. Colloid Interface Sci. 195 (1997) 368. F. Gaboriaud, J.-J. Ehrhardt, Geochim. Cosmochim. Acta 67 (2003) 967. J.R. Bargar, J.D. Kubicki, R. Reitmeyer, J.A. Davis, Geochim. Cosmochim. Acta 69 (2005) 1527. A. Naveau, F. Monteil-Rivera, J. Dumonceau, S. Boudesocque, J. Contam. Hydrol. 77 (2005) 1. W.H. Van Riemsdijk, T. Hiemstra, Adsorption to heterogeneous surfaces, in: H.E. Allen et al. (Eds.), Metals in Groundwater, Lewis Publishers, Chelsea, MI, 1993, pp. 1–36.

161

[44] T. Hiemstra, W.J.C.M. De, W.H. Van Riemsdijk, J. Colloid Interface Sci. 133 (1989) 105. [45] T. Hiemstra, W.H. Van Riemsdijk, G.H. Bolt, J. Colloid Interface Sci. 133 (1989) 91. [46] M. Villalobos, A. Perez-Gallegos, J. Colloid Interface Sci. 326 (2008) 307. [47] M. Villalobos, M.A. Cheney, J. Alcaraz-Cienfuegos, J. Colloid Interface Sci. 336 (2009) 412. [48] C. Salazar-Camacho, M. Villalobos, Geochim. Cosmochim. Acta 74 (2010) 2257. [49] T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interface Sci. 179 (1996) 488. [50] P.G. Weidler, T. Schwinn, H.E. Gaub, Clays Clay Miner. 44 (1996) 437. [51] J. Lützenkirchen, J.-F. Boily, L. Lövgren, S. Sjöberg, Geochim. Cosmochim. Acta 66 (2002) 3389. [52] T. Hiemstra, W.H. Van Riemsdijk, Geochim. Cosmochim. Acta 73 (2009) 4423. [53] J.F. Boily, J. Lutzenkirchen, O. Balmes, J. Beattie, S. Sjoberg, Colloids Surf., A 179 (2001) 11. [54] P. Venema, T. Hiemstra, P.G. Weidler, W.H. Van Riemsdijk, J. Colloid Interface Sci. 198 (1998) 282. [55] R.P.J.J. Rietra, T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interface Sci. 218 (1999) 511. [56] R.P.J.J. Rietra, T. Hiemstra, W.H. Van Riemsdijk, Environ. Sci. Technol. 35 (2001) 3369. [57] T. Hiemstra, W.H. Van Riemsdijk, Langmuir 15 (1999) 8045. [58] I.D. Brown, D. Altermatt, Acta Crystallogr., Sect. B: Struct. Sci. B41 (1985) 244. [59] C.J. Tadanier, M.J. Eick, Soil Sci. Soc. Am. J. 66 (2002) 1505. [60] T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interface Sci. 301 (2006) 1. [61] C.S. Kim, J.J. Rytuba, G.E. Brown, J. Colloid Interface Sci. 270 (2004) 9. [62] J.R. Bargar, G.E. Brown Jr., G.A. Parks, Geochim. Cosmochim. Acta 61 (1997) 2639. [63] J.R. Bargar, G.E. Brown Jr., G.A. Parks, Geochim. Cosmochim. Acta 61 (1997) 2617. [64] J.D. Ostergren, T.P. Trainor, J.R. Bargar, G.E. Brown Jr., G.A. Parks, J. Colloid Interface Sci. 225 (2000) 466. [65] J.P. Gustafsson, KTH Royal Institute of Technology, Stockholm, 2003. [66] A.L. Herbelin, J.C. Westall, Department of Chemistry, Oregon State University, Corvalis, 1999. [67] T. Hiemstra, R. Rahnemaie, W.H. Van Riemsdijk, J. Colloid Interface Sci. 278 (2004) 282. [68] R. Rahnemaie, T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interface Sci. 315 (2007) 415. [69] J.P. Gustafsson, KTH Royal Institute of Technology, Stockholm, 2011. [70] R. Rahnemaie, T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interface Sci. 293 (2006) 312. [71] D.A. Sverjensky, Geochim. Cosmochim. Acta 69 (2005) 225. [72] T. Hiemstra, M.O. Barnett, W.H. Van Riemsdijk, J. Colloid Interface Sci. 310 (2007) 8. [73] T. Hiemstra, R.P.J.J. Rietra, W.H. Van Riemsdijk, Croat. Chem. Acta 80 (2007) 313. [74] R. Rahnemaie, T. Hiemstra, W.H. Van Riemsdijk, Langmuir 23 (2007) 3680. [75] T.D. Evans, J.R. Leal, P.W. Arnold, J. Electroanal. Chem. Interfacial Electrochem. 105 (1979) 161. [76] T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interface Sci. 225 (2000) 94. [77] J. Lutzenkirchen, J.F. Boily, L. Gunneriusson, L. Lovgren, S. Sjoberg, J. Colloid Interface Sci. 317 (2008) 155. [78] M. Villalobos, M.A. Trotz, J.O. Leckie, J. Colloid Interface Sci. 268 (2003) 273. [79] R.M. Cornell, U. Schwertmann (Eds.), The Iron Oxides: Structure, Properties, Reactions, Occurrence and Uses, VCH, 1996. [80] U. Schwertmann, R.M. Cornell, Iron Oxides in the Laboratory: Preparation and Characterization, VCH, 1991. [81] U. Schwertmann, Clay Mineral. 19 (1984). [82] M. Stachowicz, T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interface Sci. 302 (2006) 62. [83] C.S. Kim, J.J. Rytuba, G.E. Brown, J. Colloid Interface Sci. 271 (2004) 1. [84] W. Schecher, D. McAvoy, Environmental Research Software, Hallowell, ME, 2003. [85] W. Schecher, Environmental Research Software, Hallowell, ME, 2001. [86] H. Wijnja, C.P. Schulthess, Soil Sci. Soc. Am. J. 65 (2001) 324. [87] A. Manceau, K.L. Nagy, L. Spadini, K.V. Ragnarsdottir, J. Colloid Interface Sci. 228 (2000) 306. [88] L. Spadini, A. Manceau, P.W. Schindler, L. Charlet, J. Colloid Interface Sci. 168 (1994) 73. [89] C.R. Collins, D.M. Sherman, K.V. Ragnarsdottir, J. Colloid Interface Sci. 219 (1999) 345. [90] S. Stokes, The University of Texas, Austin, 2009.