Carbonate adsorption on goethite in competition with phosphate

Journal of Colloid and Interface Science 315 (2007) 415–425 www.elsevier.com/locate/jcis

Carbonate adsorption on goethite in competition with phosphate Rasoul Rahnemaie 1 , Tjisse Hiemstra ∗ , Willem H. van Riemsdijk Department of Soil Quality, Wageningen University, P.O. Box 47, NL 6700 AA Wageningen, The Netherlands Received 18 May 2007; accepted 6 July 2007 Available online 7 September 2007

Abstract Competitive interaction of carbonate and phosphate on goethite has been studied quantitatively. Both anions are omnipresent in soils, sediments, and other natural systems. The PO4 –CO3 interaction has been studied in binary goethite systems containing 0–0.5 M (bi)carbonate, showing the change in the phosphate concentration as a function of pH, goethite concentration, and carbonate loading. In addition, single ion systems have been used to study carbonate adsorption as a function of pH and initial (H)CO3 concentration. The experimental data have been described with the charge distribution (CD) model. The charge distributions of the inner-sphere surface complexes of phosphate and carbonate have been calculated separately using the equilibrium geometries of the surface complexes, which have been optimized with molecular orbital calculations applying density functional theory (MO/DFT). In the CD modeling, we rely for phosphate on recent parameters from the literature. For carbonate, the surface speciation and affinity constants have been found by modeling the competitive effect of CO3 on the phosphate concentration in CO3 –PO4 systems. The CO3 constants obtained can also predict the carbonate adsorption in the absence of phosphate very well. A combination of innerand outer-sphere CO3 complexation is found. The carbonate adsorption is dominated by a bidentate inner-sphere complex, ≡(FeO)2 CO. This binuclear bidentate complex can be present in two different geometries that may have a different IR behavior. At a high PO4 and CO3 loading and a high Na+ concentration, the inner-sphere carbonate complex interacts with a Na+ ion, probably in an outer-sphere fashion. The Na+ binding constant obtained is representative of Na–carbonate complexation in solution. Outer-sphere complex formation is found to be unimportant. The 2− binding constant is comparable with the outer-sphere complexation constants of, e.g., SO2− 4 and SeO4 . © 2007 Elsevier Inc. All rights reserved. Keywords: Goethite; Hematite; Ferrihydrite; HFO; Iron oxide; ATR-FTIR; Phosphate; Carbonate; Adsorption; Extended; Stern model; CD model; MUSIC model; Competition; Speciation

1. Introduction Mineral–water interfaces play a vital role in the natural environment. Mineral surfaces bind ions, regulating their bioavailability, retention, and leaching. The binding of cations and anions on metal oxide surfaces can be described with surface complexation models. Application of surface complexation modeling to natural systems requires information on the ion binding of a substantial number of reactive elements that are omnipresent. Phosphate is a typical example. This anion is dominantly present at the surface of natural iron oxides, even if * Corresponding author. Fax: +31 317419000.

E-mail address: [email protected] (T. Hiemstra). 1 Present address: Department of Soil Science, Tarbiat Modares University,

P.O. Box 14115-336, Tehran, Iran. 0021-9797/$ – see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.jcis.2007.07.017

the concentration in solution is very low. The phosphate binding is affected by a number of macro elements and vice versa. Examples are calcium and sulfate. Calcium is an ever-present cation that interacts with phosphate in a process known as cooperative binding. The natural concentration of dissolved Ca2+ ranges from 10−5 M in some river waters [1] to 10−2 M or more in seawater [2] and soils [3]. This variation in Ca2+ affects the PO3− 4 adsorption in the pH range above 5 [4]. Another example of an omnipresent ion that interacts with phosphate is SO2− 4 . At low pH (pH < ∼5), the sulfate ion is a competitor for phosphate [5]. Sulfate is less strongly bound to Fe oxides than phosphate. However, the lower affinity is partly compensated for by a higher concentration (1 mM scale) in the environment, in particular in industrial areas with air pollution (SO2 ). In this study, we will focus on a third ubiquitous macroele. In groundwater, the main carbonate species ment, H2 CO−2+x 3

416

R. Rahnemaie et al. / Journal of Colloid and Interface Science 315 (2007) 415–425

is usually the bicarbonate ion (HCO− 3 ). At low pH, it is carbonic acid (H2 CO3 ), and at very high pH, it is carbonate (CO2− 3 ). The natural concentration range in soils and groundwater is about 10−4 –10−2 M. Higher concentrations (∼0.05–0.2 M) can be found in very alkaline environments, such as soda lakes [6]. The (bi)carbonate concentration is linked to the local partial CO2 pressure of the gas phase and the pH in solution. For goethite, the carbonate adsorption has been studied in open and closed “single ion” systems [7,8]. In closed systems, the maximum adsorption is around pH ∼ 6–7. In open systems, the carbonate adsorption increases continuously over a large pH range. For a number of ions, it has been shown that carbonate may affect the adsorption. Carbonate influences the binding of arsenate and arsenite [9], uranyl [10,11], and chromate and lead [11]. Adsorbed carbonate may also act as a competitor for phosphate. The competitive interaction between phosphate and carbonate is deliberately used in soil extraction procedures to extract phosphate [12,13]. The extraction solution of such procedures contains a very high concentration of bicarbonate, 0.5 M NaHCO3 . The mechanism of carbonate adsorption on goethite (α-FeOOH) has been studied with IR spectroscopy [14,15]. Initially, the observed IR phenomena were assigned to the formation of a mononuclear monodentate complex. Hiemstra et al. [16] argued that the data could also be interpreted as the formation of a binuclear bidentate complex. This was confirmed by Bargar et al. [17] using a combination of ATR-FTIR spectroscopy and quantum chemical calculations. In addition, Bargar et al. [17] resolved for hematite the presence of a second type of carbonate species. This species is found in the whole range of pH values studied (pH ∼ 4–8). Remarkably, it has been assigned to the adsorption of the CO2− 3 ion as an outer-sphere complex. In the present study, we will measure the PO4 –CO3 interaction for a large series of (bi)carbonate concentrations as a function of pH at different PO4 loadings. The interaction will be interpreted with the CD model [18] using the extended Stern (ES) double-layer model option [19–21]. This approach may elucidate the surface speciation of carbonate ions for the conditions studied. In the modeling, we will use for phosphate the parameter set that has recently been derived by Rahnemaie et al. [20]. The authors have obtained the ionic CD values of the relevant surface complexes from quantum-chemical optimized geometries that have been interpreted with the Brown bond valence approach [22,23]. The ionic CD values thus calculated have been corrected for the effect of water dipole orientation [19]. In our present study, the same approach will be followed to derive the CD value for adsorbed carbonate. This has the practical advantage that for the description of the CO3 –PO4 interaction, only the affinity constant of the surface complex(es) has to be adjusted. From a theoretical perspective, this approach is helpful in constraining the interpretation of adsorption data. The affinity constant(s) obtained from the CO3 –PO4 interaction will be used to predict the CO3 adsorption in single ion systems, which will be verified experimentally. We note that in this approach, the binding constants for carbonate species will be derived without any direct measurement of the carbonate ad-

sorption. The values are found by studying only the effect of carbonate on the PO4 adsorption. 2. Materials and methods 2.1. Preparation of reagents Without precautions, preparation of solutions in contact with the atmosphere will lead to dissolution of CO2 . To prevent this source of contamination, all chemicals (Merck p.a.), except carbonate solutions, were made under purified N2 and stored for a short time in polyethylene bottles to be free of silica. The acid solutions were stored in glass bottles to avoid contamination by organic materials [20]. A stock solution of NaOH was prepared CO2 -free from a highly concentrated 1:1 NaOH/H2 O. The mixture was centrifuged to remove any solid Na2 CO3 . A subsample of the supernatant was pipetted into ultrapure water and stored in a desiccator, equipped with a CO2 absorbing column. Ultrapure water (≈0.018 dS/m) was used throughout the experiments. It was preboiled to remove dissolved CO2 before using it in the experiments. The experiments were done in a constant temperature room (21 ± 1 ◦ C). 2.2. Preparation and characterization of the goethite The goethite suspension was prepared based on the method of Atkinson et al. [24], as described in detail by Hiemstra et al. [25]. Freshly prepared 0.5 M Fe(NO3 )3 was slowly titrated with 2.5 M NaOH to pH 12. The suspension was aged for 4 days at 60 ◦ C and subsequently dialyzed in ultrapure water. Before it was used in the experiments, the goethite suspension was acidified (pH ≈ 5) to desorb and remove (bi)carbonate, by continuously purging with N2 for at least one day. The BET (N2 ) specific surface area of this goethite equals 85 m2 /g. The same batch of goethite has been previously used by Rahnemaie et al. [20]. Some additional experiments were done with another batch of goethite, prepared similarly. The BET surface area of this goethite was 98.6 m2 /g. The surface charge of goethite was measured in NaNO3 solutions. A sample of goethite was titrated forward and backward by base and acid within the pH range of almost 4 to 10.5. The temperature was fixed at 20 ± 0.1 ◦ C using a thermostated reaction cell. The details of the experimental setup and data handing are given elsewhere [26]. 2.3. Carbonate–phosphate adsorption edges in binary ion systems Adsorption experiments were performed in individual gastight 23.6-ml low-density polyethylene bottles with fixed amounts of salt, goethite, carbonate, and phosphate at different pH values. All solutions, except (bi)carbonate, were added to the bottles under clean, moist N2 gas to avoid carbonate contamination. A certain amount of HNO3 or NaOH was added to the vessels, in order to obtain final pH values within the relevant pH range. After the N2 flushing was stopped at the end, the

R. Rahnemaie et al. / Journal of Colloid and Interface Science 315 (2007) 415–425

carbonate solution was added and the bottles were immediately closed. The pH range in the experiments was limited to pH > 6.5 to minimize the amount of carbonate in the gas phase of the bottle (0.18 Lgas /Lsolution ). In the modeling, we corrected by calculation for any carbonate released into the gas phase of the bottles. NaNO3 solution was used to obtain the intended concentration in the suspensions. The final volume of the suspensions was 20 ml. The bottles were equilibrated for 24 h in an end-over-end shaker in the constant-temperature room. After centrifugation, a sample of the supernatant was taken for phosphate analysis. Phosphate concentration was determined using an adapted molybdenum blue method. The pH of the suspensions was measured in the remixed suspensions. For each data point, the total concentration of components of the system was calculated based on bookkeeping of the concentrations of the added solutions. The adsorption experiments were performed at various levels of (bi)carbonate in two goethite systems (A = 85 m2 /g) with the same initial phosphate concentration (0.4 mM) but different concentrations of goethite (3 or 9 g/L). The added initial carbonate concentration was 0, 0.03, 0.1, 0.2, or 0.5 mol/L. The carbonate was added as NaHCO3 . The intended background electrolyte concentration (0.5 M) was obtained by adding the appropriate amount of NaNO3 . Above pH ∼ 8.5, the Na+ concentration increased gradually as a result of the addition of NaOH to adjust the solution pH, which has been taken into account during modeling. An additional system was prepared using another goethite (98.6 m2 /g). The intended electrolyte level in these systems was lower (0.2 M Na+ ). In these experiments, four levels of carbonate were used, i.e. 0, 0.05, 0.1, and 0.2 mol/L. The systems contained 0.4 mmol/L phosphate and 6 g/L goethite. 2.4. Carbonate adsorption edges in single ion systems To study the carbonate adsorption in the absence of any specific competitor, adsorption edges of carbonate were measured in a system without phosphate. Three levels of carbonate (1.2, 2.2, and 3.2 mmol/L) were used. The systems in the pH range 6.5–10.5 contained 10 g/L goethite and 0.1 mol/L NaNO3 . The goethite from the second batch was used (98.6 m2 /g). After equilibration and centrifugation, a sample of the supernatant was taken for analysis. The carbonate concentration in the solution phase was measured with a TOC analyzer. The TOC analyzer was adapted to measure the carbonate concentration by excluding an acidification sequence and the procedure was carefully checked by measuring a series of organic and inorganic standard solutions.

417

Table 1 Aqueous speciation reactions and their equilibrium constant (I = 0) Species

Reaction

HPO2− 4 H2 PO− 4

2− + PO3− 4 + H  HPO4 3− PO4 + 2H+  H2 PO− 4 +  H PO PO3− + 3H 3 4 4 + + H+  NaHPO− PO3− + Na 4 4 +  NaPO2− PO3− + Na 4 4 − + CO2− 3 + H  HCO3 + CO2− 3 + 2H  H2 CO3 2− CO3 + 2H+  H2 O(l) + CO2 (g) − + CO2− 3 + Na  NaCO3 2− CO3 + 2Na+  Na2 CO03 0 + + CO2− 3 + Na + H  NaHCO3 + − H + OH  H2 O Na+ + NO− 3  NaNO3

H3 PO4 NaHPO− 4 NaPO2− 4 HCO− 3 H2 CO3 CO2 NaCO− 3 Na2 CO03 NaHCO03 H2 O NaNO3 a b c d e

log K 12.35a 19.55a 21.70a 13.40b,c 2.05c 10.33a 16.69a 18.15a 1.02d 0.01a 10.14d 14.00a −0.60e

From Lindsay [54]. From Turner et al. [55]. From Rahnemaie et al. [20]. From Millero and Schreiber [56]. From Smith et al. [57].

Table 2 The charge allocation (z) and the affinity constants (log K) of ion pairs interacting with singly (=FeOH) and triply (=Fe3 O) coordinated surface groups of goethite as derived from modeling of the goethite titration data with the extended Stern (ES) model, using equal Stern layer capacitances of C1 = C2 = 0.92 ± 0.01 F/m2 [20] Ions

z0

z1

z2

log K

H+ Na+

1 0 1 1

0 1 −1 −1

0 0 0 0

9.0 −0.61 9.0 − 0.70 = 8.30 9.0 − 0.45 = 8.55

H+ –NO− 3 H+ –Cl−

model [28]. The calculated geometry has been interpreted with the Brown bond valence approach [22] to obtain the ionic charge distribution value of the CO3 surface complexes. 2.6. Surface complexation modeling The surface complexation modeling and objective optimization of the adsorption parameters have been done using the ECOSAT 4.8 program [29] in combination with a recent version (2.581) of the program FIT of Kinniburgh [30]. The important solution complexation constants are given in Table 1. 3. Results and discussion

2.5. MO/DTF computations 3.1. Primary charge The geometries of surface complexes of carbonate were optimized by molecular orbital (MO) calculations using Spartan’06 software [27]. The geometry optimizations were done using density functional theory (DFT). The basis set with pseudo potentials comprises the 6-31 + G** basis set for main group elements H–Ar. The structures were optimized with the B3lyp

Surface groups of metal (hydr)oxides may interact with electrolyte ions, forming outer-sphere complexes, known as ion pairs. These adsorbed cations and anions are located in the interface at a minimum distance of approach from the negatively or positively charged surface groups, respectively [31].

418

R. Rahnemaie et al. / Journal of Colloid and Interface Science 315 (2007) 415–425

(A)

(B)

Fig. 1. Templates with two Fe(III)–O octahedra, defined with Fe–O distances and angles as found in goethite (α-FeOOH). The central Fe(III) ions are surrounded by oxygens (large spheres) that bind one or two protons (small white spheres). Two types of oxygens are found in goethite (OI and OII ), which have a different Fe–O distance, being respectively d(Fe–OI ) ∼210 pm and d(Fe–OII ) ∼196 pm. The goethite bulk structure can be written as FeOII OI H. Template A is uncharged. The charge of template B is +2 v.u.

Fig. 2. Two Fe(III)–O octahedra with a hydrated bidentate carbonate surface complex, formed by exchange of two coordinated H2 O molecules on top of the cluster for CO2− 3 . These H2 O groups are equivalent with two singly coordinated groups at the goethite interface. The structure at the left-hand side shows the carbonate complex formed by interaction with the uncharged template A and the structure at the right-hand side is formed from the positively charged template B (see Fig. 1).

Recently, the charging behavior of goethite has been measured for a series of electrolyte ions including Li+ , Na+ , K+ , Cs+ , − − NO− 3 , Cl , and ClO4 in a consistent manner [26]. Analysis of the data indicates that the location of the head end of the diffuse double layer (DDL) differs from the minimum distance of approach of electrolyte ions. For an interpretation with all electrolyte ions in one electrostatic plane, it is crucial to separate the electrolyte ion pairs in the model from the head end of the DDL [19]. According to Hiemstra and van Riemsdijk [19], this finding can be related to the structuring of interfacial water molecules, only allowing electrolyte ions to penetrate stepwise into the last few aligned layers of water molecules near the surface; i.e., electrolyte ions occupy particular locations. The picture agrees with recent spectroscopic information. The structure of water near the surface can be studied by different methods such as force measurements [32,33] and X-ray reflectivity [34–36]. These measurements show a decaying ordering of water with distance from the surface, equivalent to about three layers of water molecules [19]. In surface complexation modeling, the alignment of water molecules pops up as the requirement of the presence of a second Stern layer in the double layer that separates the ion pairs from the head end of the DDL. This extension of the basic Stern (BS) concept is known as an extended Stern (ES) layer model [37]. In the ES model, the electrolyte ions are present in the 1-plane. The head end of the diffuse double layer (DDL) starts at another position (2-plane). The capacitance of the layer between the two electrostatic planes has been found from simultaneous modeling [20] of the consistent data set of Rahnemaie et al. [26]. The relevant parameter set is given in Table 2. The parameters of Table 2 predict the H adsorption behavior of the goethites used in this study (data are not shown).

a series of phosphate–iron complexes, using molecular orbital calculations applying density functional theory (MO/DFT). The starting point in the calculation is a cluster of two Fe oxide octahedra serving as a template to mimic the goethite mineral. The Fe–O distances, d(Fe–OI ) = 210 pm and d(Fe– OII ) = 196 pm, and the angles of the two octahedra represent the values found for goethite [38]. By adding protons to the structure, a zero-charged cluster Fe2 (OH)6 (OH2 )4 can be defined, given as template A in Fig. 1. The O–H distance is set at d(O–H) = 104 pm, as found in goethite. The location of the protons of the row of central OII H groups corresponds to the location found in goethite. In this respect, the present template differs from previous ones [19,20,39].The protons of the other row of OH groups (OI H) have the same set of bond angles as found in goethite for the coordination of the OI to the next Fe3+ . The combination of template A with a carbonate ion will result in a negatively charged moiety (−2 v.u.). This charge can be compensated for by adding two additional protons to the most basic OH ligands of the template. These OI H ions are singly coordinated and have the largest Fe–O distance (210 pm). This leads to a lower saturation of the oxygen with bond valence charge and a correspondingly higher proton affinity [40]. The resulting template (B) is given in Fig. 1. On goethite, carbonate may form bidentate inner-sphere complexes [16]. The bidentate iron–carbonate structure has been defined by exchanging both H2 O molecules on the top of the clusters for CO3 . These two exchanged ligands (Fig. 1) are equivalent to the singly coordinated surface groups at the 110 face of goethite. In order to mimic the effect of hydration, additional water molecules have been defined that interact via H bridges (O–H· · ·O) with the ligands of the CO2− 3 ion. The non2− coordinated oxygen of CO3 was allowed to interact with three water molecules via O–H· · ·H bridges. The common O ligand in both Fe–O–C bonds was allowed to interact with one H2 O. In Fig. 2, the geometry of the carbonate complex, attached to template A and template B, is given. During the optimization, the lower part of both octahedra (Fe2 (OHx )6 (OH2 )2 ) has been

3.2. Quantum chemical geometry optimizations The ionic charge distribution of a complex is related to the geometry. Rahnemaie et al. [20] have obtained the geometry of

R. Rahnemaie et al. / Journal of Colloid and Interface Science 315 (2007) 415–425

Table 3 The distances (pm) in the geometry of carbonate complexes attached to templates that are differently charged ≡(FeO)2 CO

Template Aa

Template Bb

n·H2 Oc O–Cd

2+3 130.5 130.5 128.5 210.4 ± 0.1 309.4 ± 0.1 140.5 −1.38 + 2 −0.62 + 0

2+3 131.3 132.1 126.5 207.8 ± 1.9 285.1 ± 1.2 140.5 −1.46 + 2 −0.54 + 0

O–Cd C–Od Fe–O Fe–C R0 n0 + nH0 g n1 + nH1 g

Exp. – – – 196e – 139.0f

Note. The carbonate ion is hydrated with water molecules (n·H2 O), coordinating to respectively the common, and the free oxygen ligand. The geometry is optimized with the MO/DFT-B3lyp model. a Template A is uncharged but the combination with CO2− leads to a nega3 tively charged (−2 v.u.) moiety. b Template B has two additional protons. In combination with the CO2− unit, 3 it is uncharged. c The number of water molecules coordinating with respectively the common oxygen(s) and the free ligand(s) of the complexes. d O–C refers to the bond of C with the common oxygen and C–O refers to the bond with the free O ligand. e Distance present in the goethite structure without relaxation. f Average R for carbonate in minerals [22]. 0 g The n and n values represent the partial charge of CO2− (n + n = 0 1 0 1 3 −2 v.u.) attributed to the 0- and 1-planes. These coefficients are calculated combining the Brown bond valences and the charge of the oxygens ligands placed in the electrostatic plane. The nH0 and nH1 values represent the charge of additional protons that are located in respectively the 0- and 1-planes.

fixed. All bonds in the hydrated (Fe–O)2 –C–O–(H2 O)5 moiety (upper half of the geometry) were allowed to relax. As is obvious from Fig. 2, the presence of additional protons on the iron octahedra has a considerable effect on the geometry of the iron–carbonate complex. Apparently, the angle of the Fe–O–C bond is flexible. In the case of binding to template B (Fig. 2b), the moiety is bent. This is due to interaction of the free, negatively charged O ligand of the CO3 ion with the positive charges on both Fe–OI H1/2 2 groups of template B. Without this charge (template A), the carbonate bends away (Fig. 2a). Previously, we have used template A with a different orientation of the protons present in the row with OI H ions. The carbonate complex formed with that template [19] is almost identical to the present results found for template A. This supports the above explanation that charge on the ligands can be considered a key factor in the geometry of the carbonate complex. The adsorption of CO3 mainly occurs in the pH range below the PZC of goethite. In that case, the surface will be positively charged. Therefore, we consider the CO3 geometry interacting with template B as most representative for adsorbed CO3 . This will be discussed later in more detail. The geometry of our bidentate iron–carbonate complexes can be compared with the geometry calculated at full relaxation of the Fe octahedra as given by Bargar et al. [17]. They showed the geometry of a positively charged cluster, 2− Fe2 (OH)4 (OH2 )2+ 6 , that has reacted with a CO3 ion, releasing two water molecules, which are present as hydration water. The (FeO)2 CO moiety is not bent and in this respect, it is compara-

419

ble to our results obtained with template A, although the charge of their cluster is comparable to the charge of template B. In their calculation with full relaxation, the ligands of the Fe cluster have moved, whereas in our case (template B) the C–O has strongly moved. 3.3. Ionic CD values The optimized geometries (Table 3) have been used to calculate the ionic charge distribution of the adsorbed carbonate ion, applying the Brown bond valence concept [22]. In this concept, the bond valence s and distance R are related, according to s = e−(R−R0 )/B ,

(1)

in which B is a constant (B = 37 pm) and R0 is the element specific parameter. The value of R0 is chosen so that the sum of the bond valences around the C ion corresponds to the formal C valence (z = 4+). Application of Eq. (1) results in the charge distribution of the carbonate ion (n0 , n1 ). These numbers can be combined with the charge of the adsorbing protons to calculate the ionic charge distribution coefficients (n0 + nH0 , n1 + nH1 ) for an adsorption reaction. For the adsorption reaction, forming a bidentate surface complex we write ≡2FeOH−1/2 + 2H+ (aq) + CO2− 3 (aq) −1+z0 z1 CO + 2H2 O, log K(FeO)2 CO .  ≡(FeO)2

(2)

The value of nH0 is equal to nH0 = 2, because two protons are adsorbed by the ≡FeOH−1/2 groups, forming ≡FeOH1/2 2 . Both water groups are exchanged for oxygens of the carbonate ion, leading to a certain charge attribution of the carbonate ion to the surface (n0 ). No protons are bound to the free ligand of the adsorbed carbonate; i.e., nH1 = 0. The final ionic charge distribution is given by n0 + nH0 and n1 + nH1 . The sum of the charges is equal to the charge added by the two protons and the carbonate ion (Eq. (2)); i.e., n0 + nH0 + n1 + nH1 = 0. The ionic charge distribution thus calculated is given in Table 3. Our calculations show that the CDs of the two bidentate carbonate complexes are slightly different. The numbers found for the complex formed on template A are almost equal to the value previously reported [19] for a very comparable template (n0 + nH0 = 0.62 v.u., n1 + nH1 = −0.62 v.u.). In the case of interaction of CO3 with a positively charged template (B), the bond lengths change, which leads to a lower surface charge attribution (n0 + nH0 = 0.54 v.u.). 3.4. Electrostatic dipole corrections Sum frequency spectroscopy is a method of studying interfacial water [41,42]. It shows that the number of polar-oriented water molecules increases if the interface becomes charged [41]. Moreover, it shows that the dipole orientation may change on either side of the PZC [43]. The orientation of dipoles near the surface can be interpreted as electrostatic feedback on the introduction of charge at the surface [19]. In the CD model, the dipole orientation effect is included in the value of the interfacial charge distribution coefficients

420

R. Rahnemaie et al. / Journal of Colloid and Interface Science 315 (2007) 415–425

Species

n0 + nH0 a

n1 + nH1 a

z0

z1

low and intermediate PO4 loading. At low pH and a relatively high PO4 loading, phosphate is also adsorbed as a protonated monodentate (MH) complex. In the modeling of the PO4 –CO3 interaction, we will use the phosphate adsorption parameters (Table 5) derived by Rahnemaie et al. [20].

≡(FeO)2 CO (A) ≡(FeO)2 CO (B)

0.62 ± 0.01 0.54 ± 0.01

−0.62 −0.54

0.68 0.62

−0.68 −0.62

3.6. Phosphate–carbonate interaction

Table 4 The ionic charge distribution values (n0 + nH0 , n1 + nH1 ) and corresponding overall charge distribution coefficients (z0 , z1 ) for the carbonate complexes formed with templates A and B

a The ionic CD values given refer to the average value found with the

MO/DFT model options (B3lyp) used in this study (Table 3). The CD values refer to reaction Eq. (2). Table 5 Tableaux defining the reactions of phosphate and protons with singly coordinated surface groups forming surface complexes Species

≡FeOH−1/2

H+

PO3− 4

z0

z1

z2

log K

≡(FeO)2 PO2 ≡FeOPO2 OH

2 1

2 2

1 1

0.46 0.28

−1.46 −1.28

0 0

29.72 27.63

Note. The geometry-based CD values and the log K values, optimized on adsorption data, are from Rahnemaie et al. [20].

(z0 , z1 ). Therefore, the overall CD values may differ from the above calculated ionic charge distribution n0 +nH0 and n1 + nH1 (Table 3). The interfacial and ionic charge distributions are related according to [19]    z0 = n0 + nH0 − φ n0 + nH0 + (3) nref zref and

   z1 = n1 + nH1 + φ n0 + nH0 + nref zref ,

(4)

in which nH0 and nH1 are the numbers of protons that are located in respectively the 0- and 1-plane according to the formation reaction that uses a number of reference groups nref with a corresponding charge zref . The factor φ is a proportionality constant (φ ≈ 0.17 ± 0.03). The calculated dipole correction for the bidentate complex is about 0.08 v.u., i.e. small but significant. The calculated interfacial charge distribution coefficients (z0 , z1 ) are given in Table 4 for both binuclear bidentate carbonate complexes.

In Fig. 3, the experimental phosphate adsorption data are given as a function of pH and carbonate concentration for three phosphate loadings. The data of Figs. 3a–3c refer to three levels of goethite concentrations, respectively 3, 6, and 9 g/L. This results in three different levels of phosphate (respectively 1.5, 1.0, and 0.5 µmol/m2 ). The systems are prepared at a constant Na+ concentration (see materials and methods), which is 0.5 molar in Figs. 3a and 3c, and 0.2 molar in Fig. 3b. The data cover a wide CO3 concentration range. High CO3 concentrations are used because PO4 is an extremely good competitor for carbonate. The data are presented and evaluated, focusing on the concentration in solution. Scaling and evaluation of the adsorption density may easily lead to situations in which modeling is less critical, because of the relatively low variation in adsorption density with changing conditions in comparison to the variation of the concentration in solution. Moreover, in many applications, the concentration is the parameter to be predicted. The experimental data show that the phosphate concentration in solution increases with increase of the level of added carbonate. The carbonate ions compete with phosphate for surface sites, leading to a decrease in the adsorption of PO4 when CO3 is added. The figure also shows a larger response to the addition of carbonate at a lower PO4 level (compare Figs. 3a and 3c). The maximum interaction between carbonate and phosphate ions takes place in the lowest part of the pH range of our study. The competitive interaction of CO3 decreases as the pH increases and is very weak above pH ≈ 10.5. These observations agree qualitatively with the adsorption behavior of CO3 in closed systems, as shown by Van Geen et al. [7] and Villalobos and Leckie [8]. Their experiments for closed systems show a maximum CO3 binding at a pH of about 6–7.

3.5. Surface chemistry of phosphate

3.7. CD modeling

As discussed in Rahnemaie et al. [20], the phosphate surface complexation on goethite has been studied experimentally with in situ CIR-FTIR [44] and theoretically with MO/DFT calculations [45]. According to Tejedor-Tejedor and Anderson [44], the main surface phosphate species identified by FTIR is a bidentate surface complex (B). This complex is protonated at low pH (BH). Kwon and Kubicki [45] suggested the presence of a deprotonated bidentate complex (BH2 ) at low pH, formation of either a deprotonated bidentate (B) or a monoprotonated monodentate (MH) complex at neutral pH, and a deprotonated monodentate (M) complex at very high pH. Recently, Rahnemaie et al. [20] have used the CD model to derive the PO4 surface speciation. For a series of possible surface complexes, the CD was independently calculated from the geometry. Application to an extended data set showed that phosphate is mainly adsorbed as a bidentate (B) complex at

The mechanism of carbonate binding to goethite has been studied with IR spectroscopy [14,15,47,48]. Symmetry is important in IR spectroscopy. Metal coordination to CO2− 3 will lead to band splitting. The experimental band splitting for carbonate bound to goethite is found to be about ν ≈ 150–200 cm−1 [14,15,48,49]. Initially, this relatively small band splitting was considered as evidence for the formation of a monodentate surface complex. However, interpretation as a hydrated binuclear–bidentate complex, i.e., =(FeO)2 CO· · ·(H2 O)n , is also possible since the band splitting can be lowered by H-bond formation in the case of hydration of the complex [16,50]. Recently, the extent of band splitting of this hydrated binuclear bidentate complex has been assessed with MO/DFT computations [17]. A small band spitting has been found (ν ≈ 190 cm−1 ), confirming our interpretation as binuclear bidentate complex formation.

R. Rahnemaie et al. / Journal of Colloid and Interface Science 315 (2007) 415–425

(a)

showing that indeed about 2/3 of the charge of CO2− 3 is shared with the surface [15,16]. In D2 O, the presence of a band at about 1460 cm−1 has been reported for goethite [14] and for hematite [17]. This band is most significant in the absence of background electrolyte ions [17]. For goethite [14], the band is found in a 1 mM NaDCO3 solution without additional electrolyte. According to Bargar et al. [17], this band is due to adsorbed CO2− 3 , present as an outer-sphere complex in the interface, as evidenced by the agreement with MO/DFT calculations showing the 1460 cm−1 wavenumber for a hydrated outer-sphere CO2− 3 complex. No evidence is found for the formation of a HCO− 3 outer-sphere complex, although this species is predominantly present in the solution in the pH range studied. Based on the above, we have defined, in addition to the formation of a bidentate inner-sphere complex (Eq. (2)), the adsorption of carbonate as an outer-sphere complex according to ≡FeOH−1/2 + H+ (aq) + CO2− 3 (aq) 2− 1/2  ≡FeOH2 · · · CO3 , log KFeOH2 –CO3 ,

(5a)

≡Fe3 O + H (aq) + CO2− 3 (aq)  ≡Fe3 OH1/2 · · · CO2− , log KFe3 OH–CO3 . 3

(5b)

−1/2

(b)

(c) Fig. 3. Experimental data and model description of phosphate interaction with goethite as a function of pH for different concentrations of carbonate (closed systems) and two electrolyte levels. The total amount of phosphate is 0.4 mmol/L. The goethite concentration is 3, 6, and 9 g/L in parts a, b, and c, respectively, which leads to different phosphate concentrations. The lines are the equilibrium concentrations of phosphate calculated with the CD model, using the extended Stern model as double layer option. The phosphate surface speciation and electrolyte parameters (Tables 2 and 5) are from Rahnemaie et al. [20]. The affinity constants of adsorbed carbonate complexes were optimized on the data (Table 6).

Evidence for the formation of a bidentate surface can also come from the interpretation of macroscopic adsorption data [16]. In the case of bidentate complex formation, 2/3 of the ligands of the CO3 ion are shared with the surface, while in case of a monodentate complex it is only 1/3. This difference is large and can be resolved with surface complexation modeling

421

+

In the modeling, the full charge of the carbonate ion has been attributed to the 1-plane. The model description of the PO4 –CO3 interactions is given in Fig. 3 (lines), showing the phosphate equilibrium concentration as a function of pH for various carbonate levels in goethite systems with different PO4 loadings. In a first model approach, we assumed the formation of a combination of inner and outer-sphere complexes (Eqs. (2) and (5a) and (5b)). This combination is able to describe the PO4 –CO3 competition data well (R 2 = 0.97) at a relatively low PO4 loading (0.5 µmol/m2 ). However, at a high PO4 loading of, for instance, 1.5 µmol/m2 (Fig. 3a), the experimental PO4 concentrations are clearly higher than predicted with the parameters optimized at the low PO4 level (not shown); i.e., the predicted PO4 adsorption is too high. This may point to the presence of an additional carbonate complex that is competing with the PO4 species. This additional species has to be relatively important at a high PO4 loading. The description of the data can be improved substantially by including a carbonate species that adds positive charges to the 1-plane due to interaction with a Na+ ion. Formation of a Na+ carbonate complex has been postulated previously by Villalobos and Leckie [15]. We propose that the Na+ ion forms an outer-sphere complex with the inner-sphere bidentate carbonate surface complex resulting in ≡(FeO)2 CO· · ·Na+ . According to Bargar et al. [17], such a sodium–carbonate surface complex is probably indistinguishable in the IR spectrum from the surface complex lacking this interaction. The formation reaction can be formulated as 2≡FeOH−1/2 + 2H+ (aq) + Na+ (aq) + CO2− 3 (aq) −1+z0

 ≡(FeO)2

log K(FeO)2 CO···Na .

CO · · · Na+1+z1 + 2H2 O(l), (6)

422

R. Rahnemaie et al. / Journal of Colloid and Interface Science 315 (2007) 415–425

Table 6 The charge allocation used and fitted log K values describing the CO3 interaction at the surface of goethite (N = 241 data points, R 2 = 0.988) Species

z0

z1

z2

log K a

Eq.

≡(FeO)2 CO ≡(FeO)2 CO· · ·Na+ 2− ≡FeOH1/2 2 · · ·CO3

0.62b 0.62c 1

−0.62 −0.62 + 1 −2

0 0 0

22.01 ± 0.02 22.03 ± 0.05 10.22 ± 0.20d

(2) (6) (5a)

≡Fe3 OH1/2 · · ·CO2− 3

1

−2

0

10.22 ± 0.20d

(5b)

Note. In the formation reactions, ≡FeOH−1/2 and ≡Fe3 O−1/2 , H+ , Na+ , and CO2− 3 are assumed to be the reference species. a The logK values have been optimized on the PO –CO interaction data. 4 3 b These CD values have been calculated for template B (Table 4). c The charge distribution in the adsorbed carbonate ion is set equal to the CD value calculated for ≡(FeO)2 CO. d The affinity constant of both outer-sphere carbonate complexes is set equal (Eq. (5)). Note that the inner-sphere complexes do not react with Fe3 O sites since this would lead to oversaturation of the common oxygen charge [16].

In the modeling, we assume for both inner-sphere complexes, i.e., ≡(FeO)2 CO and ≡(FeO)2 CO· · ·Na+ , the same distribution of charge for the carbonate ion. The introduction of the sodium carbonate complex leads to a very good description of the data (Fig. 3). An overview of the charge allocation parameters used and the fitted affinity constants is given in Table 6. It is interesting to focus for a moment on the affinity constants derived (Table 6). The difference in log K values of the two inner-sphere complexes ( log K = 0.0 ± 0.07) can be interpreted as the affinity constant of the Na+ ion for the ≡(FeO)2 CO surface complex, according to the reaction ≡(FeO)2 CO + Na+ (aq)  ≡(FeO)2 CO· · ·Na+ . The log K value can be compared with the ion pair formation constants of Na+ with (bi)carbonate species in solution (Table 1). The ion pair formation constants of Na+ (aq) with HCO− 3 (aq), 2− (aq), and CO (aq) are respectively log K ≈ −0.2, NaCO− 3 3 −1.0, and 1.0. These log K values are around the log K value obtained in our modeling for the association of Na+ with the carbonate surface complex (log K ≈ 0.0). The Na+ ion pair formation constant is also comparable to the values found for the Na+ interaction with surface sites of oxides. For instance, for goethite [20], gibbsite [51], and rutile/anatase [52], the reported Na+ ion binding constants are respectively log K = −0.6, 0.2, and −0.6. The second ion pair formation constant derived in this study is for carbonate. The affinity constant for the reaction 2− 1/2 + CO2− is log K = ≡FeOH1/2 2 3 (aq)  ≡FeOH2 · · ·CO3 1.2 ± 0.2. The log K value is clearly higher than found for monovalent anions on goethite [20], like NO− 3 (log K ≈ −0.7) and Cl− (log K ≈ −0.4), but is of the same order as found 2− for SO2− 4 and SeO4 (log K ≈ 1–2) [46,53]. It is important to notice that our modeling could not reveal the presence of any HCO− 3 as outer-sphere complexes if its presence is allowed in the modeling of our competition systems. In the above approach, we have used the calculated CD values as a constraint in the modeling. In an ideal situation, the surface speciation with corresponding charge distribution can be derived from the adsorption data. However, data sets are often too limited to allow such an approach, in particular if more than one species can be present. In the present situation,

Fig. 4. The carbonate equilibrium concentrations in single ion systems, closed to the atmosphere (10 g goethite/L, A = 98.6 m2 /g, 0.1 M NaNO3 ) as a function of pH and added total carbonate levels. The lines are pure prediction using the affinity constants (Table 6) derived from modeling the PO4 concentration in PO4 –CO3 systems (Fig. 3).

a free fit of the CD value of the carbonate complex leads to a value most closely to the one found with for the optimized geometry with template B as basis. The fitted CD values are z0 = 0.55 ± 0.04 v.u. and z1 = −0.55 ± 0.04 v.u. if in the modeling the CD value of carbonate ion present in ≡(FeO)2 CO and ≡(FeO)2 CO· · ·Na+ is kept equal. 3.8. Adsorption of carbonate in single ion systems The binding properties of CO3 have been derived indirectly from the PO4 –CO3 binary system. In modeling, the affinities of adsorbed inner-sphere and outer-sphere surface complexes of CO3 were optimized on PO4 adsorption data. The reliability depends on the PO4 surface speciation used. Therefore, the adsorption of carbonate in single ion systems has been measured to examine whether the affinities derived (Table 6) are capable of predicting carbonate adsorption in single ion systems, at a much lower carbonate loading in a background electrolyte of 0.1 M NaNO3 . In Fig. 4, the experimental data of the carbonate interaction with goethite are given as a function of pH and initial carbonate concentration. With an increase of the pH, the equilibrium concentrations increase, due to lower adsorption by goethite. The calculated lines are pure model predictions using the parameters derived by fitting the PO4 concentration data of the PO4 –CO3 competition systems (Table 6). Agreement between the prediction and the experiment is good. Note again that the parameters of Table 6 have been derived without measuring directly the adsorption of carbonate ion in the competition systems, giving confidence in the reliability of the approach. The data of the single ion system (Fig. 4) can also be used derive the affinity constant of the dominant surface species, i.e., ≡Fe2 O2 CO. The fit (R 2 = 0.94) results in almost the same log K value (21.89 ± 0.07) as found with the approach in which we did not at all measure the carbonate adsorption, but where we relied on the CD model, using only the effect of CO3 on the PO4 adsorption.

R. Rahnemaie et al. / Journal of Colloid and Interface Science 315 (2007) 415–425

423

(a)

(a)

(b)

(b) Fig. 5. The pH-dependent adsorption of carbonate onto goethite in single ion systems that are closed (a) or open (b) to CO2 in 0.01 M NaCl (closed symbols) or 0.1 M NaCl (open symbols). The data are from Villalobos and Leckie [8]. The lines are predictions using the affinity constants (Table 6) derived from modeling the PO4 concentration in PO4 –CO3 systems (Fig. 3).

The affinity constants of Table 6 have also been used to test the ability to describe the extensive carbonate data set of Villalobos and Leckie [8]. The carbonate adsorption in their open and closed systems could be described rather well (R 2 = 0.97) as shown in Fig. 5. 3.9. Carbonate surface speciation As shown above, the carbonate adsorption parameters are able to describe the carbonate behavior in single ion and dual adsorbate systems. The surface speciation chosen is based on spectroscopic information and is constrained by the independently derived charge distribution values of these surface species using the MO/DFT optimized geometry. We have calculated the surface speciation for some of the systems used in the present study (Fig. 6). In Fig. 6a, the carbonate speciation is shown for a closed single ion system as used in one of our experiments (Fig. 4). The total carbonate concentration is 1 mM Na2−x Hx CO3 at a background electrolyte level of 0.1 M for 10 g/L goethite. The

Fig. 6. (a) Surface speciation of carbonate in a closed single ion goethite system (10 g/L, A = 98.6 m2 /g) with 1 mM CO3 in 0.1 M NaNO3 . (b) Surface speciation of carbonate in closed PO4 –CO3 goethite system (9 g/L, A = 85 m2 /g) with 0.4 mM PO4 and 0.5 M CO3 /NO3 . The lines are calculated using Tables 1, 2, 5, and 6.

calculation shows dominance of the ≡(FeO)2 CO species. Outer sphere complexation is not very important. Maybe surprisingly, the ≡(FeO)2 CO· · ·Na+ species is also not relevant, even though 0.1 M NaNO3 is present. This situation changes in systems such as the one given in Fig. 6b. In the figure, the carbonate speciation is shown for a PO4 –CO3 system (9 g/L goethite) with 0.4 mM PO4 (∼0.5 µmol/m2 ). The total carbonate concentration Na2−x Hx CO3 is 0.5 M. In this system, the PO4 adsorption adds a considerable amount of negative charge to the interface, which stimulates the formation of ≡(FeO)2 CO· · ·Na+ in combination with a very high Na+ concentration in solution. The high carbonate concentration leads to a relatively high CO3 surface loading. For this system, a small amount of some outersphere complexes is calculated, present at high pH. So far, the interpretation of the carbonate adsorption has been straightforward, leading to a rational surface speciation with a consistent set of parameters. However, one observation remains to be explained. At a very low ionic strength, the ATRFTIR spectrum of hematite shows two sets of bands when equilibrated with 361 ppmV CO2 in D2 O at various pD values. One set of bands (1330 and 1530 cm−1 ) is assigned to a binuclear bidentate complex. The other set (1350 and 1460 cm−1 ) is attributed to the presence of CO2− 3 as an outer-sphere complex. These bands are also found for goethite in 1 mM Dx CO−2+x 3 solution at pD = 4.6 and 7.0. However, when we calculate the

424

R. Rahnemaie et al. / Journal of Colloid and Interface Science 315 (2007) 415–425

expected surface speciation in 1 mM NaHx CO−2+x solution, 3 only the binuclear bidentate complex (≡(FeO)2 CO) is found to be present over the whole range of pH values. The outer-sphere complexation is extremely low. If we simulate the conditions used in the constant P–CO2 experiment of Bargar et al. [17], we come to the same conclusion. If we assume that our calculations are essentially correct, we need an explanation for the presence of two species in the IR spectrum. According to our MO/DFT calculations (Fig. 3), carbonate can be present in two geometries. Both species differ in bending, which is due to a different interaction with charged surface groups. Besides bending, both species have also a different set of C–O bond lengths (Table 3). These differences may lead to a difference in the vibrational behavior. At present, the extent of this difference is unknown. The simultaneous presence of both types of carbonate species can be simulated, assuming that both species has approximately the same intrinsic affinity. However, the relative presence will depend on conditions like pH, ion loading, and ionic strength. The relative presence is regulated by the difference in the CD value of both types of complexes. A different set of CD values will lead to a different response of the species in the electrostatic field radiated by the surface. The combination of the electrostatic potentials of the 0- and 1-plane is relevant. The surface plane will have a positive potential at pH < PZC. In the presence of adsorbed PO4 , the particle will become negatively charged, with a negative electrostatic potential in the 1-plane. This will favor the bidentate species that adds the smallest amount of negative charge to the 1-plane and the highest amount of negative charge the surface plane. The expected carbonate surface species under these conditions (a high CO3 and/or PO4 loading) will be the species with an ionic CD value of n0 = −1.46 + 2, n1 = −0.54 v.u., i.e., ≡(FeO)2 CO (B). In the absence of adsorbed PO4 and at a low carbonate loading, the electrostatic preference for this species is reduced, which may lead to the presence of both species. This combination will be most marked in the absence of electrolyte ions because these ions will suppress the potential in the 1-plane via ion pair formation. It is interesting to note that a free fit of the carbonate adsorption in the single ion systems of Villalobos and Leckie [8] leads to z0 = 0.70 ± 0.03 v.u. [19], a value closest to template A. In contrast, a free fit of the CD value for the present PO4 –CO3 data leads to z0 = 0.55 ± 0.04 v.u., which is closer to template B. The above-sketched situation can also be translated to a molecular picture. In the absence of electrolytes ions, the carbonate species of template A may be relatively important (Fig. 2a). On the average, the CO3 ions are bent more away from the surface. Accumulation of negative charge near the surface, for instance due to binding of PO4 , will lead to a bending of the CO3 molecule toward the positively charged surface groups, forming the carbonate species as calculated with template B. Although we have chosen to use the latter species in our modeling of the CO3 –PO4 data, the data can also be described very well by a combination of two bidentate complexes that differ slightly in the CD value. However, it is difficult to pinpoint at present the relative affinities of the species involved.

4. Conclusions The above can be summarized in a series of conclusions: • Carbonate interactions with goethite can be described with the CD model using a combination of inner- and outersphere complexation. • The main inner-sphere species is a binuclear bidentate species, ≡(FeO)2 CO. This species may interact with Na+ in an outer-sphere fashion, forming ≡(FeO)2 CO· · ·Na+ . This process is stimulated by the adsorption of PO4 . • Carbonate outer-sphere complexation is usually very limited. Any outer-sphere complexation is due to interaction of carbonate rather than bicarbonate with protonated sur2− 2− 1/2 face groups (FeOH1/2 2 · · ·CO3 and Fe3 OH · · ·CO3 ). • The carbonate adsorption of goethite can be predicted without a direct measurement of the adsorption in single ion systems, analyzing only the effect of carbonate on the PO4 adsorption with an ion complexation model that correctly describes the competitive interaction. • The MO/DFT optimized geometry of the binuclear bidentate CO3 complex depends on the charge of the template used. A positively charged template leads to bending of the carbonate complex and a larger attribution of negative charge to the common ligands. These complexes will behave differently in an electrostatic field. Therefore, the relative presence is regulated by macroscopic factors like pH, electrolyte level, and ion loading. • Carbonate ions bind much more weakly than phosphate at the goethite surface. They are nevertheless able to act as competitors for PO4 if carbonate is present in a relatively high concentration. Acknowledgments We appreciate the financial support of the Ministry of Science, Research, and Technology of Iran (MSRT). The work was partly funded by the EU project, FUNMIG (516514, F16W2004). The authors thank Mr. A. Korteweg (Laboratory of Physical and Colloid Chemistry) for the BET analysis. References [1] M.F. Benedetti, S. Mounier, N. Filizola, J. Benaim, P. Seyler, Hydrol. Process. 17 (2003) 1363. [2] R. Chester, Marine Geochemistry, second ed., Blackwell Science, Oxford, 2003. [3] G. Sposito, The Thermodynamics of Soil Solutions, Oxford Univ. Press, New York, 1981. [4] R. Rietra, T. Hiemstra, W.H. Van Riemsdijk, Environ. Sci. Technol. 35 (2001) 3369. [5] J.S. Geelhoed, T. Hiemstra, W.H. Van Riemsdijk, Geochim. Cosmochim. Acta 61 (1997) 2389. [6] E. Kebede, Z.G. Mariam, I. Ahlgre, Hydrobiologia 288 (1994) 1. [7] A. Van Geen, A.P. Robertson, J.O. Leckie, Geochim. Cosmochim. Acta 58 (1994) 2073. [8] M. Villalobos, J.O. Leckie, Geochim. Cosmochim. Acta 64 (2000) 3787. [9] M. Stachowicz, T. Hiemstra, W.H. Van Riemsdijk, Environ. Sci. Technol. (2007), doi: 10.1021/es063087i.

R. Rahnemaie et al. / Journal of Colloid and Interface Science 315 (2007) 415–425

[10] M. Wazne, G.P. Korfiatis, X.G. Meng, Environ. Sci. Technol. 37 (2003) 3619. [11] M. Villalobos, M.A. Trotz, J.O. Leckie, Environ. Sci. Technol. 35 (2001) 3849. [12] S.R. Olsen, C.V. Cole, F. Watanabe, L.A. Dean, in: United States Department of Agriculture, vol. 939, Washington, DC, 1954, p. 1. [13] N.J. Barrow, T.C. Shaw, Geoderma 16 (1976) 109. [14] H. Wijnja, C.P. Schulthess, Soil Sci. Soc. Am. J. 65 (2000) 324. [15] M. Villalobos, J.O. Leckie, J. Colloid Interface Sci. 235 (2001) 15. [16] T. Hiemstra, R. Rahnemaie, W.H. Van Riemsdijk, J. Colloid Interface Sci. 278 (2004) 282. [17] J.R. Bargar, J.D. Kubicki, R. Reitmeyer, J.A. Davis, Geochim. Cosmochim. Acta 69 (2005) 1527. [18] T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interface Sci. 179 (1996) 488. [19] T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interface Sci. 301 (2006) 1. [20] R. Rahnemaie, T. Hiemstra, W.H. Van Riemsdijk, Langmuir 23 (2007) 3680. [21] T. Hiemstra, M.O. Barnett, W.H. Van Riemsdijk, J. Colloid Interface Sci. 310 (2007) 8. [22] I.D. Brown, D. Altermatt, Acta Cryst. B 41 (1985) 244. [23] I.D. Brown, The Chemical Bond in Inorganic Chemistry: The Bond Valence Model, Oxford Univ. Press, New York, 2002, p. 12. [24] R.J. Atkinson, A.M. Posner, J.P. Quirk, J. Phys. Chem. 71 (1967) 550. [25] T. Hiemstra, J.C.M. De Wit, W.H. Van Riemsdijk, J. Colloid Interface Sci. 133 (1989) 105. [26] R. Rahnemaie, T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interface Sci. 293 (2006) 312. [27] Anonymous Spartan ’06 Wavefunction Inc., Irvine, CA, USA, 2006. [28] J. Kong, C.A. White, A.I. Krylov, C.D. Sherrill, R.D. Adamson, T.R. Furlani, M.S. Lee, A.M. Lee, S.R. Gwaltney, T.R. Adams, C. Ochsenfeld, A.T.B. Gilbert, G.S. Kedziora, V.A. Rassolov, D.R. Maurice, N. Nair, Y. Shao, N.A. Besley, P.E. Maslen, J.P. Dombroski, H. Daschel, W. Zhang, P.P. Korambath, J. Baker, E.F.C. Byrd, T. Van Voorhis, M. Oumi, S. Hirata, C.-P. Hsu, N. Ishikawa, J. Florian, A. Warshel, B.G. Johnson, P.M.W. Gill, M. Head-Gordon, J.A. Pople, J. Comput. Chem. 21 (2000) 1532. [29] M.G. Keizer, W.H. Van Riemsdijk, ECOSAT, Technical Report Department Soil Science and Plant Nutrition, Wageningen Agricultural University, Wageningen, The Netherlands, 1998. [30] D.G. Kinniburgh, Fit, Technical Report WD/93/23, British Geological Survey, Keyworth, Great Britain, 1993.

425

[31] O. Stern, Z. Electrochem. 30 (1924) 508. [32] R.M. Pashley, J.N. Israelachvili, J. Colloid Interface Sci. 101 (1984) 511. [33] J.N. Israelachvili, H. Wennerstrom, Nature 379 (1996) 219. [34] M.L. Schlegel, K.L. Nagy, P. Fenter, L. Cheng, N.C. Sturchio, S.D. Jacobsen, Geochim. Cosmochim. Acta 70 (2006) 3549. [35] P. Fenter, N.C. Sturchio, Prog. Surf. Sci. 77 (2004) 171. [36] J.G. Catalano, C. Park, Z. Zhang, P. Fenter, Langmuir 22 (2006) 4668. [37] J. Westall, H. Hohl, Adv. Colloid Interface Sci. 12 (1980) 265. [38] J.-L. Hazemann, J.F. Berar, A. Manceau, Mater. Sci. Forum 79 (1991) 821. [39] T. Hiemstra, M.O. Barnett, W.H. Van Riemsdijk, J. Colloid Interface Sci. 310 (2007) 8. [40] T. Hiemstra, P. Venema, W.H. Van Riemsdijk, J. Colloid Interface Sci. 184 (1996) 680. [41] V. Ostroverkhov, G.A. Waychunas, Y.R. Shen, Phys. Rev. Lett. 94 (2005). [42] Y.R. Shen, V. Ostroverkhov, Chem. Rev. 106 (2006) 1140. [43] M.S. Yeganeh, S.M. Dougal, H.S. Pink, Phys. Rev. Lett. 83 (1999) 1179. [44] M.I. Tejedor-Tejedor, M.A. Anderson, Langmuir 6 (1990) 602. [45] K.D. Kwon, J.D. Kubicki, Langmuir 20 (2004) 9249. [46] R. Rahnemaie, T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interface Sci. 297 (2006) 379. [47] J.D. Ostergren, T.P. Trainor, J.R. Bargar, G.E. Brown Jr., G.A. Parks, J. Colloid Interface Sci. 225 (2000) 466. [48] W.A. Zeltner, M.A. Anderson, Langmuir 4 (1988) 469. [49] J.D. Ostergren, J.R. Bargar, G.E. Brown, G.A. Parks, J. Synch. Radiat. 6 (1999) 645. [50] J.P. Jolivet, Y. Thomas, B. Taravel, V. Lorenzelli, G. Busca, J. Mol. Struct. 79 (1982) 403. [51] T. Hiemstra, H. Yong, W.H. Van Riemsdijk, Langmuir 15 (1999) 5942. [52] K. Bourikas, T. Hiemstra, W.H. Van Riemsdijk, Langmuir 17 (2001) 749. [53] R. Rietra, T. Hiemstra, W.H. Van Riemsdijk, J. Colloid Interface Sci. 240 (2001) 384. [54] W.L. Lindsay, Chemical Equilibria in Soils, Wiley–Interscience, New York, 1979. [55] D.R. Turner, M. Whitfield, A.G. Dickson, Geochim. Cosmochim. Acta 45 (6) (1981) 855. [56] F.J. Millero, D.R. Schreiber, Am. J. Sci. 282 (9) (1982) 1508. [57] R.M. Smith, A.E. Martell, Critical Stability Constants, Plenum, New York, 1981.