Measuring and Removing Dissolved Metals from Stormwater in Highly Urbanized Areas (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

11 3.1 Introduction 3.1.1 Metal Speciation The fate and transport of metals in natural surface waters is highly dependent on the properties of the particular metal, the solution chemistry of the water (i.e., pH, ionic strength, redox state, and presence of biotic, organic, and inorganic ligands), and interactions with resident particulate matter in the sys- tem. Figure 3-1 provides a schematic of the different potential reactions that affect metal ion speciation in water. Water chemistry parameters such as pH, metal ion con- centration, the presence of other reactive ligands and metals, ionic strength, and redox potential dictate metal ion specia- tion within the water column through complexation and oxidation/reduction processes. These processes impact the extent and rates of interaction with particulate matter and the bioavailability of metals. Interactions with particulate matter include sorption to organic and inorganic phases as well as surface mediated transformation reactions (e.g., redox reactions). The mineral- ogy and available SA of solid phases in the systems, in conjunc- tion with water chemistry parameters such as pH and ionic strength, dictate the rates and extent of adsorption. Sorption processes, as well as precipitation reactions, often control the rates and extent of migration and removal of metals in surface water, groundwater, stormwater, and highway runoff as well as the bioavailability of metal ions. The extent of sorption and precipitation are dependent on aqueous metal ion speciation such that metal ion speciation in the aqueous phase and sorp- tion processes to inorganic, organic, and biological particulate matter are coupled. Finally, colloids, which are particles small enough to pass through conventional filters used to separate dissolved and suspended phases, are major carriers of trace elements. As a result, sorption of trace elements to colloidal matter, includ- ing nanoparticles, can have a significant impact on metal ion mobility. This chapter reviews the key properties that control the chemical speciation of metals within water, as well as the key processes and parameters that affect the rate and extent of removal of metal ions from particulate matter. A detailed understanding of these processes will not only facilitate pre- dicting the fate, transport, and bioavailability of metal ions in natural water and stormwater, but will also guide the devel- opment of treatment strategies. Parameters to be explored within the literature include those associated with the metals (e.g., potential for adsorption to mineral surfaces and organic matter, oxidation/reduction, complexation, and precipita- tion) as well as properties of the water, potential ligands, and particulate matter surfaces. While the focus of this task will be on speciation in natural fresh waters, it is important to recog- nize that the processes that affect speciation in natural waters are the same as those that dictate speciation in stormwater and highway runoff. The major difference lies in the composition of the background water and particulate phases. 3.1.2 Natural Water Composition Table 3-1 presents a summary of the concentrations of major elements in polluted and unpolluted river water (Maurice 2009). There are large variations of Ca2+, HCO3-, and Na+ that reflect differences in bedrock chemistry. For example, the presence of limestone (CaCO3) in bedrock increases the hardness (sum of the multivalent cations) and alkalinity of the water. Trace elements, defined as those present at concentrations less than 1 mg/L, constitute the remaining elements. Because their concentrations are low, they are not typically accounted for in measures of TDS. However, it is important to recognize that some of these other elements such as iron, aluminum, and titanium may constitute significant fractions of underly- ing bedrock. Aluminum, iron, and titanium have very limited mobility due to their solubility; however, their presence in the sediment layer often impacts the mobility of other trace elements. A compilation of data by Gaillardet et al. (2005) C H A P T E R 3 Environmental Chemistry of Metals in Surface Waters

12 summarizes the concentration of trace elements in major rivers across the world. Average concentrations of selected elements are presented in Table 3-2. Figure 3-2 depicts the wide variability among the differ- ent elements as well as the several order of magnitude span in concentration for each element. Correlations among different elements have also been studied. For example, by normalizing trace metal concentrations to continental crust composition, Gaillardet et al. (2005) identified average global mobility trends. Highly mobile elements were defined to have mobilities close to or greater than that of sodium. Included in this group are chlorine, carbon, sulfur, cadmium, boron, selenium, arsenic, calcium, magnesium, and strontium. The second group having mobilities approximately 10 times less than sodium includes silicon, lithium, potassium, manganese, barium, copper, cobalt, and nickel. The third group, referred to as the non-mobile group because their mobilities are 10–100 times less than sodium, includes zinc, chromium, vanadium, lead, and iron. The final group is defined by mobility indexes more than 100 times less than sodium and includes aluminum and titanium. Thus, the heavy metals cadmium, copper, and zinc are expected to have varying mobilities in natural waters based on this analysis. These researchers (Gaillardet et al. 1999) utilized this data to develop a relationship between the fraction of an element in the dissolved phase based on the linear equilibrium partitioning coefficient, Kd (mg/gm of metal ion sorbed per Figure 3-1. Speciation of metal ions in natural waters (adapted from Zaporozhets et al. 2010). Metal species ColloidsSuspended Dissolved Complexed Inorganic Organic Radionuclides Organometallic compounds (e.g. Methylmercury) Ions in varying oxidations states (e.g. Cu+1, Cu+2) Bound to organic particles Within sulfide, carbonate, and aluminosilicate minerals Adsorbed on minerals (e.g. iron, silica, aluminum) With inorganic ligands (e.g. lead chloride, copper carbonate) With high-molecular organic ligands (i.e. humic substances, proteins) With low-molecular organic ligands (i.e. aminoacids, carboxylic acids, EDTA, etc.) Ca2+ Mg2+ Na+ K+ Cl- SO42- HCO3- SiO2 World actual 0.36 0.15 0.31 0.03 0.23 0.12 0.87 0.18 World unpolluted 0.33 0.14 0.22 0.03 0.16 0.09 0.85 0.17 Table 3-1. Concentrations (in mmol/L) of major elements in world average “actual” and “natural (unpolluted)” river water (Maurice 2009). Al As Cd Cr Cu Fe Mn Ni Pb Zn 32 0.62 0.08 0.7 1.48 66 34 0.801 0.079 0.6 Table 3-2. Average concentrations (in mg/L) of selected trace elements in world river water (Gaillardet et al. 2005).

13 mg/L of metal ion in solution), and the TSS concentration. Their results are summarized in Figure 3-3 and suggest that for the world average TSS concentration of 350 mg/L, only the most mobile elements (e.g., Na, B, Se, As) will be transported in the dissolved phase. Other correlations for estimating sorbed and dissolved frac- tions of metal ion have also been conducted based on cluster analysis or water quality parameters. Seyler and Boaventura (2001) reported that dissolved concentrations of vanadium, copper, arsenic, barium, and uranium are strongly positively correlated with major elements and pH. pH and major ele- ment composition can be affected by anthropogenic inputs depending on the natural bedrock composition. Thus, bedrock composition, chemical properties, and large scale processes such as acid rain can impact on metal ion concentrations within a river. 3.1.3 Characterization of Natural Organic Matter in Receiving Waters The fate of metal ions in water is also highly dependent on the presence of natural organic matter that is ubiquitous in natural waters. Typical concentrations of DOM range from less than 2 to nearly 10 mg/L as C in freshwater streams and can be significantly higher in lakes (see Table 3-3). However, values as high as 80 mg/L have been reported for rivers drain- ing wetlands (Walther 2005). While DOM is often quantified in terms of the concentration of carbon, the complexity and Figure 3-2. Selected trace element concentrations in world river waters (adapted from Gaillardet et al. 2005). Figure 3-3. Predicted dissolved metal ion transport as a function of suspended solid concentration in river water (adapted from Gaillardet et al. 2005).

14 heterogeneity of natural organic matter has significant impacts on its reactivity. The organic matter found in natural aquatic systems is a complex mixture of partially decomposed vegetation, animal matter, and other decaying parent material. As Figure 3-4 sug- gests, the composition of organic matter is made up of a range of chemical moieties including both soluble and insoluble frac- tions, carbohydrates, proteins, fats, and complex organic macromolecules (Abdulla et al. 2010). These substances range in size from low molecular weight dissolved substances to insoluble particulate matter (see Figure 3-5). The three main sources of DOM, such as fresh plant litter, microbial biomass, and humus, vary with respect to climate, season, and geology. Humic substances (shown in the left side of Figure 3-5) formed during the decomposition of vegetation, animal matter, and other decaying parent material represent the most impor- tant components of aquatic DOM as they are present at signifi- cant concentrations in natural waters. The chemistry of their formation is quite complex and originates from both terrestrial origin (autocthonous sources) and biological activity within Source DOC (mg/L as C) Humic Substances (mg/L as C) Sea water 0.2 - 2.0 0.06 - 0.6 Groundwater 0.1 - 2.0 0.03- 0.6 River 1 - 10 0.5 - 4.0 Lake 1 -50 0.5 - 40 Table 3-3. Typical ranges of DOC concentrations and humic substance concentrations in natural waters (Thurman 1985). Figure 3-4. Formation of natural organic matter (adapted from Grunwald 1998).

15 the natural water (allocthanous sources). However, despite the wide variety of sources and decomposition pathways, there is significant uniformity in the macroscopic properties of humic matter. Thus, a significant amount of early research focused on generalizing these gross properties and developing classifica- tion schemes that were based on these properties. For example, elemental analysis indicates that humic substances contain car- bon, oxygen, hydrogen, nitrogen, and sulfur. Forty to 50% of the mass is comprised of complex aliphatic carbon chains (e.g., C-C-C-C-) and 35% to 60% of the mass is 4, 5, and 6 member carbon rings (termed aromatic components due to the alternat- ing double bonds) with C-C, C-N, and C=O components. One of the most widely known classification schemes for humic substances is based on differences in solubility of vari- ous components of humic substances. The three major com- ponents, humin, humic, and fulvic acid, vary with respect to solubility in alkali and acidic solutions as shown in Figure 3-6. These different fractions vary with respect to color, composi- tion, and reactivity. Humins are black in color and insoluble at any pH. In addition to improving water-holding capacity and improving soil structure, humins function as a cation exchange system in soils. Thus, it is not surprising that substances con- taining significant fractions of humins (e.g., peat) have been used as adsorbents for metal ions (Ringqvist and Oborn 2002; Liu et al. 2008; Shareef 2009). Humic substances are ambiguous with respect to their affinity for water. They contain a hydrophobic (lipophilic) backbone that is made up of non-polar hydrocarbon bonds and hydrophilic, polar functional groups containing oxygen, nitrogen, and sulfur atoms. Thus, they are often referred to as hydrophobic acids. Non-humic and fulvic acid organic mat- ter is often characterized by their hydrophobic and hydrophilic components as well. This type of classification is reflected in Figure 3-7, which shows that the relative distribution of organic matter varies from groundwater to surface water, and the ratio of humic to fulvic acid is slightly higher in lake and river water. The acidic nature of the humic and fulvic acid polar, organic groups suggests that their presence can influence the pH and buffering capacity of natural water (Köhler et al. 1999). In addition, these functional groups provide binding or com- plexation sites for metal ions. The association of metal ions with humic substances has been shown to influence the rates of sorption and precipitation (Scheinost et al. 2001; Lin et al. 2005; Zhao et al. 2010). As a result, researchers have sought to more clearly elucidate the structure of humic substances and to quantify the number and reactivity of the various types of functional groups in humic substances. Even though significant research has been conducted over the past several decades to characterize the properties Figure 3-5. Distribution of sizes of various components of organic matter (adapted from Steinburg 2003). Figure 3-6. Operationally defined classification of humic substances (adapted from Steinburg 2003).

16 and reactivity of humic substances, there is still a lack of consensus as to the nature of these complex materials (Hayes and Clapp 2001; Perminova et al. 2009). Researchers gener- ally agree that much of the reactivity of humic and fulvic acids is due in large part to the presence of carboxylic (-COOH) and phenolic (C6H5OH) groups (Steinburg 2003). Carboxylic groups are much more acidic (Ka’s on the order of 10-4.5) than phenolic groups (Ka’s on the order of 10-10). Thus, fulvic acids are expected to be more reactive than humic acids because they are higher in carboxylic acids and lower in phenolic content. These differences are evident in Table 3-4 and shown schemati- cally in Figure 3-8 and Figure 3-9. The composition of organic matter and relative percentages of humic and fulvic acids var- ies depending on the origin of the organic matter. Terrestrial humic substances contain mainly lignoprotein (lignin + pro- tein) complexes and humic and fulvic acids are major constitu- ents. In contrast, aquatic humic substances are predominantly carbohydrate protein complexes composed mostly of fulvic acids. Thus, aquatic humic substances exhibit acidity to water, which is especially important in waters low in other natural buffers such as carbonate and bicarbonate ions (Köhler et al. 2001; Koopal et al. 1999). Structural models of humic substances vary throughout the literature. Fulvic acids are a mixture of weak aliphatic and aro- matic organic acids, but their shape and composition varies. In contrast, humic substances have traditionally been represented as flexible linear polymers that exist as random coils with cross- linked bonds to micelles. The most current views assume either that the structure of humic substances is macropolymeric with relatively large organic molecules considered as flexible linear synthetic polyelectrolytes (Figures 3-8 and 3-9), or humic sub- stances are supramolecular associations of relatively small, chemically diverse organic molecules that have self-organized into large macromolecules. It is now thought that biomolecu- lar moieties are also associated with these substances (Sutton and Sposito 2005). Hydrophobic Neutrals 4% Hydrophilic Acids 24% Hydrophilic Bases 3% Hydrophilic Neutrals 17% Humic Acids 8% Fulvic Acids 44% Hydrophobic Acids 52% Lake and River Water Hydrophobic Neutrals 4% Hydrophilic Acids 50% Hydrophilic Bases 2% Hydrophilic Neutrals 30% Humic Acids 1% Fulvic Acids 13% Hydrophobic Acids 14% Groundwater Figure 3-7. Distribution of hydrophilic and hydrophobic fractions of natural organic matter (adapted from Peuravuori and Pihlaja 2000).

17 Regardless of which model of humic substances emerges from the on-going inquiry, it is clear that these humic sub- stances provide functional groups and hydrophobic moieties that can interact with surfaces and dissolved metals ions in solution. These interactions can significantly impact the mobility and toxicity of metal ions in water. 3.1.4 Summary Average concentrations of trace elements in freshwaters, stormwater, and highway runoff range over orders of mag- nitude, and their mobility in these systems vary depending on the metal ion, the chemical speciation, and the extent of Constituent Humic Acid Fulvic Acid pKa C 54-59 41-51 H 3.2-6.2 3.8-7.0 N 0.8-4.3 0.9-3.3 O 33-38 40-50 S 0.1-1.5 0.1-3.6 O/C 0.51 0.74 H/C 1.0 1.4 total acid functional groups 5.6-8.9 41-51 carboxylic acids 1.5-5.7 5.2-11 6-9 typically deprotonated in circumneutral pH range phenolic OH 2.1-5.7 0.3-5.7 9 alcoholic OH 0.2-4.9 2.6-9.5 14 increases solubility quinoide/keto, C=O 0.1-5.6 0.3-3.1 increases water solubility slightly methoxy, OCH3 0.3-0.8 0.3-1.2 Table 3-4. Elemental composition (weight %), mole ratios, and functional group content (meq/g) of humic and fulvic acids (Steinburg 2003). Figure 3-8. Structure of humic acid proposed by Stevenson (1982). Figure 3-9. Structure of humic acid proposed by Buffle et al. (1977).

18 partitioning. For the average suspended solid concentration of 350 mg/L, only the most mobile elements (e.g., Na, B, Se, As) will be transported in the dissolved phase. As a result, parti- tioning to particulate matter and colloids in natural water is a key component to understanding fate and transport in water. Anthropogenic and natural organic matter also has a large impact on metal ion transport. The composition of natu- ral organic matter is extremely complex and heterogeneous; however, the high reactivity and solubility of humic and ful- vic acids suggests that these fractions are the key components for interaction with metal ions in natural waters. The research suggests that much of the reactivity of humic and fulvic acids is due in large part to the presence of carboxylic (-COOH) and phenolic (C6H5OH) groups. These same assumptions may not be true for anthropogenic inputs as they tend to be dominated by hydrophobic and lower molecular weight compounds. 3.2 Metal Ion Speciation in Solution: Complexation Reactions Metal ions in solution are always associated with other ions, organic species, or particulate matter. Their coordination envi- ronment is dictated by the composition of the water including the pH, ionic strength, and the presence of other ions, organic molecules, and solid phases. Complexation reactions, in which a metal ion forms an association or bond with another ion (ligand), are the dominant reaction that controls metal ion speciation in water. Indeed, even in the absence of other mol- ecules in the water, metal ions form complexes with the water molecules themselves, forming a primary hydration sphere around the metal ion. 3.2.1 Complexation with Inorganic Ligands Metal ion speciation often dictates the mobility and tox- icity in a particular aquatic environment because complex- ation reactions prevent metal ions from being precipitated, complexing agents act as carriers for trace elements in water, and metal ion bioavailability is often reduced by complex- ation. Metal ion speciation often involves the formation of a bond between a central metal ion and inorganic anions or functional groups that have electron pairs to donate. In natural waters, there are a number of simple inorganic ligands such as OH-, Cl-, NH3, CO3 2-, and PO4 3-, as well as natural and anthropogenic organic ligands that may be available within the system to bind to a central metal ion (e.g., Cu, Fe, etc.) to form a soluble complex as shown in Figure 3-10. The importance of complexation in natural waters depends on the pH of the water and the concentration and binding strength of available ligands. Stability constants describing the binding strength for the reaction between a particular metal ion and the common inorganic ligands are well estab- CH3 O O OH2 OH2 OH2 OH2 OH2 Cu+ C Monodentate O O OOH2 OH2 OH2 OH2 Cu C O OOOH2 OH2 OH2 OH2 Cu O C C Bidentate (Cu acetate)+ CuCO30 [Cu oxalate]0 CH2 OOOH2 OH2 O OH2 Cu- O C C C O CH2 CH2 C O OH Muldentate [Cu citrate] O H OH O HH O Fe- O H2 O H2 Fe2- N NH O O O O N N H Muldentate, polynuclear Figure 3-10. Examples of several types of metal-ligand complexes.

19 lished in the literature and maintained in NIST databases (Martell and Smith 1974). Complexation reactions with simple monodentate, inorganic ligands such as those repre- sented by the reactions of Hg, Pb, Cu, and Zn with the chloride ligand are shown in Table 3-5 and highlight the variability in binding strength for different metals. The value of the stability constant, b value, for these reac- tions represents the equilibrium constant for the multi-step reaction of each metal ion with the associated number of chloride ligands. Note that the divalent metals can bond with more than one chloride ligand, but only one chloride ligand bonds with the metal ion. Comparison of the different metals suggests that the strength of the metal-ligand complex follows the trend: Hg>Pb>Cu>Zn. Nevertheless, all of these binding constants are relatively weak and typical of ion pair associations with the exception of the HgCln metal-ligand constants that dis- play more covalent character. The low values of these sta- bility constants indicate that chloride complexes of Pb, Cu, and Zn will not significantly affect metal ion speciation in freshwaters. In contrast, carbonate complexes range from approximately 105 to 1010.5 for these same metals, and their concentrations in solution can be significant as shown in Figure 3-11. When the hydroxide ion is the dominant ligand in the sys- tem, many metal ions act as multiprotic acids in which pro- tons are released from water molecules in the inner hydration sphere as shown in Figure 3-12. Reaction β β βReaction Reaction Hg2+ + Cl- HgCl+ 107.2 Hg2+ +2 Cl- HgCl2 1014 Hg2+ +3Cl- HgCl3 1015.1 Pb2+ + Cl- PbCl+ 101.6 Pb 2+ +2 Cl- PbCl2 101.8 Pb 2+ +3Cl- PbCl3 101.7 Cu2+ + Cl- CuCl+ 100.5 Zn2+ + Cl- ZnCl+ 100.4 Zn 2+ +2 Cl- ZnCl2 100.2 Zn 2+ +3Cl- ZnCl3 100.5 Table 3-5. Stability constants (b) for selected chloride complexes. 0 10 20 30 40 50 60 70 80 90 100 Pe rc en to fT ot al Aq ue ou s D is tr ib uti on pH 6 pH 7.5 0 10 20 30 40 50 60 70 80 90 100 Pe rc en t o f T ot al A qu eo us D is tr ib uti on pH 6 pH 7.5 0 10 20 30 40 50 60 70 80 90 100 Pe rc en to fT ot al Aq ue ou s Di st rib uti on pH 6 pH 7.5 0 10 20 30 40 50 60 70 80 90 100 Pe rc en to fT ot al Aq ue ou s D is tr ib uti on pH 6 pH 7.5 Figure 3-11. Impact of carbonate and sulfate complexes on metal ion speciation in solutions calculated using Visual MINTEQ and typical freshwater ligand concentrations.

20 For example, the first hydrolysis reaction of Fe+3 [written with its waters of hydration as Fe(H2O)6 3+] with the hydroxide ligand can be written as either a complexation reaction: Fe H O + OH Fe H O OH + H O =102 6 3 2 5 2 2 11.8( ) ( ) ( )⇔ β+ − + or an acid/base reaction, Fe H O Fe H O OH H Ka = =102 6 3 2 5 2 2.2( ) ( ) ( )⇔ + ∗β+ + + − where the  indicates that the stability constant is written in terms of H+ rather than OH-. As pH is increased, hydroxide ligands can replace additional water molecules (i.e., more hydro- gen ions are released) to sequentially form Fe(H2O)4(OH)2+, Fe(H2O)3(OH)3 0, and Fe(H2O)2(OH)4-. The value of Ka = b = 10-2.2 suggests that Fe(H2O)63+ is a relatively strong acid. Lead, copper, zinc, and cadmium are significantly weaker acids with acidity equilibrium constants (Ka values) for their first proton dissociation of 10-7.71, 10-8.0, 10-8.96, and 10-10.0, respectively, compared to the value of 10-2.2 noted for Fe(III) above. Thus, Fe(III) acts as an acid that is as strong as phosphoric acid (H3PO4), whereas the acidity of zinc and cadmium are only slightly more acidic than bicarbonate (HCO3-). Fig ure 3-13 further demonstrates that the impor- tance of hydroxide complexes on speciation is dependent on the particular metal ion. The reactivity of metal cations in aqueous systems is often correlated with the values of b1 as shown earlier for the trends in complexation with chloride and ethylendiamine tetraacetate (EDTA). Of particular interest with respect to mobility is that complexation of metal ions has a signifi- cant effect on the charge of the species. For example, metal hydroxide complexes can be cations [e.g.,Cu(OH)+], neutral molecules [e.g., Cu(OH)2 0, (aq)], or anions [e.g., Cu(OH)3-]. The relative importance or dominance of each of these charged species is pH dependent. For example, speciation of Cr+3 and Cu2+ as a function of pH shown in Figure 3-14 and Figure 3-15, respectively, demonstrates the change in species charge as more hydroxide ligands or carbonate ligands are available for complexation. Indeed, at pH values below 4, chromium speciation is dominated by the neutral species and at pH values above 7, a negatively charged hydroxide complex dominates. For Cu2+, the presence of 10-3M of carbonate impacts the speciation at circumneutral pH values where the neutral by charged CuCO3o complex dominates. The distribution of charge affects both the OH2 OH2 OH2 OH2 H O H OH2 Fe3+ OH2 OH2 OH2 OH2 OH- OH2 Fe3+ + H + Figure 3-12. Representation of the acid/base properties of Fe3+ in water in octahedral coordination with water and hydroxyl ligands. 2 3 4 5 6 7 8 9 10 11 12 Mn Fe Co Cu Zn Pb Cd pH Me(OH)2 Me(OH) Me2+ Figure 3-13. Distribution of metal hydroxide complexes as a function of pH (adapted from Essington 2005).

21 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Lo g C (M ) pH Cr +3 Cr(OH) +2 Cr(OH)2 + Cr(OH)3 Cr(OH)4 - Figure 3-14. Speciation diagrams for 10-3 M Cr(III) in water at 25°C in the absence of competing ligands calculated using Visual MINTEQ. -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Lo g C (M ) pH Cu 2+ CuOH + Cu(OH)2 Cu(OH)3 - Cu(OH)4 -2 CuHCO3 + CuCO3 Cu(CO3)2 - Figure 3-15. Speciation of 10-3 M Cu(II) in water at 25°C in the presence of 10-3M total carbonate calculated using Visual MINTEQ.

22 adsorptive properties of the metal and transport across biotic membranes. 3.2.2 Complexation in the Presence of Chelating Agents In many natural systems multidentate ligands or chelat- ing agents are present. These types of species are also rep- resented in Table 3-6. These chelating agents form stronger bonds with the central metal ion because they donate several pairs of electrons. Many of the chelating agents identified in natural waters are from anthropogenic sources. For example, the most well-known complexing agents are the aminopoly- carboxylates, of which EDTA is a well-known representative. As shown in Table 3-6, the trend in binding strengths for the metal-EDTA complexes is similar to the chloride complexes; however, the binding strengths (as reflected in the value of the equilibrium constants) are orders of magnitude greater than those reported for the metal-chloride complexes. As a result, EDTA is mostly found bound to metal ions in natural waters. The values of these stability constants suggest that when EDTA is added to water at concentrations comparable to or greater than the metal ions, EDTA-metal complexes will domi- nate the metal ion speciation and these complexes will carry a net negative charge. Natural chelating agents are also found in surface waters and are often generated from biological processes. Ethylenediamine-succinate (EDSA) is a metabolite of the soil bacteria, Amycolatopsis orientalis. In contrast to EDTA, EDSA is readily biodegradable. Siderophores such as desferrioxamine-B (DFO-B) and azotochelin are naturally produced chelating agents that bind trace metals including Cu(II), Cd(II), and Pb(II) (Bellenger et al. 2007; Mishra et al. 2009). Complexation in these systems is a function of metal ion and pH. For example, Mishra et al. (2009) found minimal complexation of DFO-B with Pb at pH 3.0, but at pH 7.5 to 9.0, Pb was bound to three hydroxamate groups of the ligand in a “caged” hexadentate structure. At intermediate pH values (pH 4.8), a mixture of Pb-DFO-B complexes involving bind- ing of the metal through one and two hydroxamate groups was observed. In contrast, Cd remained dissociated at pH 5.0, was present as a mixture of Cd-DFO-B and inorganic (hydroxide) species at pH 8.0, and was bound by three hydroxamate groups from DFO-B only at the highest pH value, pH = 9.0. Stability constants for azotochelin and various metals are shown in Table 3-7. Of the metals shown in Table 3-7, copper exhibits the strongest affinity for azotochelin. Of particular interest is that this siderophore has recently been found to contain complex metals that form oxyanions such as molyb- date, vanadate, and tungstate (Bellenger et al. 2007). Oxyani- ons are polyatomic ions that contain a metal ion and several oxygen atoms such as phosphate, sulfate, arsenate, and molyb- date. They are relatively mobile in water due to their negative charge and relatively large in size compared to cationic metal ions such as Fe(III), Cu(II) and Zn(II). Complexation can have a significant impact on metal bio- availability. In some cases, the presence of ligands shields the charge of the metal ion and allows it to pass across metal cell membranes. In other cases, the size of the metal-ligand cluster sterically hinders membrane transport (Finney and O’Halloran 2003; Manceau and Matynia 2010). Complexation also has a significant impact on metal ion solubility, and chelating agents are often used to extract metal ions from soils, sediments, and precipitated solids. 3.2.3 Complexation by Humic Substances Humic and fulvic acids also serve as chelating agents for metal ions because they contain hydroxyl, phenoxylic, and carboxylic functional groups as shown in Figure 3-8 and Figure 3-9. As with other complexing agents, humic substances interact with and change the solubility and toxicity of trace elements in natu- ral aqueous environments. In addition, major cations such as Na+, Mg2+, and Ca2+ can bind to organic matter to promote aggregation of smaller, dissolved natural organic molecules into colloidal particles. Due to the heterogeneity of these molecules, Hg2+ + EDTA4- HgEDTA2- = 1023.5β Pb2+ + EDTA4- PbEDTA2- = 1019.8 β Cu2+ + EDTA4- CuEDTA2- = 1020.5 β Zn2+ + EDTA4- ZnEDTA2- = 1018.3β Table 3-6. Stability constants for selected EDTA-metal complexes. Metal Ion Reaction Log of the Stability Constant, log β Co2+ Co2+ + LH5 CoLH2- + 4H+ Log β1 = -23.0 Co2+ + LH5 CoLH2- + 3H+ Log β1 = -13.5 Cu2+ Cu2+ + LH5 CuLH2- + 4H+ Log β1 = -12.9 Zn2+ Zn2+ + LH5 ZnL3- + 5H+ Log β1 = -24.1 Zn2+ + LH5 ZnLH2- + 4H+ Log β1 = -17.83 Table 3-7. Complexation reactions of various metal ions with azotochelin, a naturally occurring siderophore (adapted from Bellenger et al. 2007).

23 humic and fulvic acid binding is typically characterized using a range of ligand sites that vary with respect to their binding strength. Many researchers have employed small, monodentate organic acids such as those shown in Figure 3-16 to characterize weak binding of metals to fulvic acid, and multidentate ligands such as phthalate, salicylate, citrate, and oxy-succinate to quan- tify the strong binding sites (Leenheer et al. 1998; Nantsis and Carper 1998; Smith and Kramer 2000; Otto et al. 2001). The stability constants (b values) for different metal– humic acid complexes typically follow a common order: Cu>Fe>Pb>Ni>Co>Ca>Cd>Zn>Mn>Mg. Note that this trend is not consistent with the previous order of reactivity shown for the inorganic ligands and anthropogenic chelating agents described earlier. Most notable is the relative stability of Cu-humic complexes. Manceau and Matynia (2010) used x-ray absorption spectroscopy to suggest that the stability of Cu-humic complexes is due to the abundance of dicarboxyl- ate sites on the aliphatic portions of the molecule which pro- vide good fits to the geometry of complexed Cu(II). These spectroscopic results are supported by work of Meylan et al. (2004) in which in situ measurements of labile and dissolved copper and zinc in natural freshwater were compared to predictions from several models for the com- plexation by humic and fulvic substances. In all cases, the prediction of Zn2+ and complexed zinc concentrations were in agreement with the experimental data. However, model over-predictions of free Cu2+ concentrations were attrib- uted to strong specific copper-binding ligands that were not accounted for in the models. This enhanced affinity for Cu suggests that the presence of humic acids in natural water will increase the overall solubility of Cu in water as both free Cu2+ and bound Cu will be present in solution. Hoffmann et al. (2007) determined binding constants for copper and zinc with colloidal and dissolved organic ligands in river water on the order of b = 1013. Equilibrium constants for the ligand-metal complex formation constants from ultra- filtered samples less than 3,000 Daltons were about 0.5 log units stronger than the values obtained from the 0.4 micron filterable fraction, suggesting the importance of colloidal humic matter. Nonetheless, the strength of these complexes is less than similar complexes formed with EDTA. Indeed, addi- tion of EDTA to a solution containing metal organic matter complexes, led to dissociation of the complexes and forma- tion of metal-EDTA complexes (Schmitt and Frimmel 2003). While general trends in complexation with natural organic matter (NOM) have been observed, complexation with NOM is more difficult to define due to the complex and ill-defined nature of these organic complexes and the variability in the number and types of functional groups. Because the func- tional groups associated with humic substances are respon- sible for binding of humic and fulvic acids to mineral surfaces and metal ions, identifying the binding reactions with metal ions is difficult because (1) humic substances have different numbers of binding sites, (2) site competition is prevalent, (3) stoichiometry is metal ion dependent, and (4) electrostatics influence the extent of binding. In most studies, carboxylic acid and phenolic functional groups have dominated the speciation with metals. The relative importance of each is a function of the nature of the metal ion and the acidity of the func- tional group. In general, complexation of the carboxylic acid H OH2 OH2 OH2 OH2 M OH2 O H O O C H H H C C C C C C R1 R2 O C H H H C C C C C C R1 R2 OH2 OH2 OH2 OH2 M OH2 O Salicylatemodel (bidentate: carboxylate and phenolate) O C OH2 H H C C C C C C R1 R2 OH2 OH2 OH2 O M O Phthalate model (bidentate: o dicarboxylate) O OH2 O C O H H C C C C C C R1 R2 OH2 OH2 OH2 M C O Figure 3-16. Organic molecules used to characterize metal ion binding to fulvic acid (adapted from Leenheer et al. 1998 and Essington 2005).

24 and phenolic functional groups with metal cations increases with pH, decreases with metal ion concentration, and increases with humic acid concentration. Recent studies have suggested that acid polysaccharides participate in metal binding in col- loidal organic matter and that binding of metals to humic acids greater than 1000 molecular weight units (Daltons) is associ- ated with colloidal humic material depending on the metal ion and solution properties (Stordal et al. 1996; Quigley et al. 2002). Researchers have also emphasized the presence of sulfur moieties in humic substances, and have suggested that binding to sulfur sites may be important for adsorption of many metal ions (Smith et al. 2002; Karlsson et al. 2005). Even with this complexity, reasonable agreement for binding of particular metals to humic substances extracted from natural waters frequently yields similar binding properties. For exam- ple, Cheng and Allen (2002) studied zinc binding to organic matter isolates (obtained using reverse osmosis) to isolate the organic matter. Using anodic stripping voltammetry and iden- tical pH, ionic strength, and DOC concentrations, they found that the zinc titration curves from the different surface water sources were similar enough to suggest that the binding con- stants among the different waters were similar. They also dem- onstrated that the zinc-NOM complexes become more stable at higher pH. In contrast, Groenenberg et al. (2010) demonstrated that metal-fulvic acid binding parameters, optimized through modeling experimental data, showed substantial variation. Monte Carlo simulations showed differences in uncertain- ties depending on the affinity of the metal for the fulvic acids. For high affinity metals such as Cu and Pb, variations among samples were largely due to the natural variation in the log b values. However, for metals with a lower affinity such as Cd, uncertainty was associated with estimation of the fraction of fulvic acid in the organic matter. Competition among different metal ions for complexation with humic and fulvic acids is also complicated due to the heterogeneous nature of these substances. While trace metal complexes such as Cu-NOM and Pb-NOM are generally stronger than major metal ion complexes such as Ca-NOM and Mg-NOM complexes, the concentration of major ions in natural waters makes them effective competitors for NOM complexation. Competition between trace metals and major ions is feasible even at these different concentration levels if the metal ions compete for the same sites on the organic matter. However, researchers have postulated that different ions pref- erentially complex to different functional groups of the humic and fulvic acids. For example, researchers have postulated that Cu complexation in natural waters occurs at the phenolic sites (Lu and Allen 2006) and phenolic and carboxylic functional groups (Benedetti et al. 1995). They attributed differences in pH dependency for Ca-NOM and Cu-NOM to the dominance of calcium complexation to the carboxylic sites, especially at high Ca concentration and low pH. Iglesias et al. (2003) also suggested that calcium only competes for specific copper sites. This research also suggested that only a fraction of the proton active sites are involved with Cu complexation. Other researchers have focused on speciation to colloidal organic matter. Burba et al. (2001) developed a conceptual model based on the assumption that naturally occurring macromolecules and colloids ranging in size from nano- meters to microns and comprised of irregular networks of organic and inorganic moieties are enriched in Al, Cu, Mn, Pb, and Zn. These so-called hydrocolloids compete for metal ions with organic matter and external ligands (e.g., EDTA and organic matter for Cu) and competition for organic mat- ter occurs between different metal ions (e.g., Cu exchanges for Mn, Mg and Ca and Al and Fe exchange for Cu). These results suggest that models for complexation of metal ions to humic and fulvic acids must incorporate multiple com- plexation sites, account for pH dependency associated with proton release from the sites, and include major ion and ligand competition. Two of the most challenging tasks for developing models to predict humic-metal ion competition are estimat- ing the distribution and number of these different ligand sites and determining the stability (or equilibrium) constants for the formation of metal-ligand complexes. 3.2.4 Modeling Metal Complexation A number of different models have emerged that describe the acid-base behavior of natural waters and metal ion binding to inorganic and organic complexes. The speciation of metal ions in waters containing simple inorganic ligands, anthropo- genic chelating agents such as EDTA, and simple organic acids is well established and numerous chemical speciation models (PHREEQC, MINEQL, MINTEQA2, GEOCHEM) are adept at predicting the species distribution in systems containing metal ions and these ligands. However, the major challenge associated with speciation in natural waters has been describ- ing and predicting chemical speciation in waters containing natural and anthropogenic organic matter. Such models must be able to predict metal ion binding over a wide range of con- ditions as a function of pH, concentration of organic matter, ionic strength, and the presence of competing metal ions and ligands. For example, each model must include a value for the total concentration of binding sites for each ligand, and a corresponding stability constant for the formation of each metal-ligand complex. The model must also describe how protons bind and dissociate by including reactions and equi- librium constants for each acid and base in the system. Thus, a complete model incorporates the equilibrium constants for each binding reaction and each proton dissociation reaction, mass balances for each type of ligand and each site on a ligand (e.g., phenolic and carboxylic if only two sites are used), mass balances for each metal ion, and a charge balance.

25 Over the past several decades several modeling approaches have evolved that have shown potential for predicting metal ion binding to NOM in these complex systems. The key to develop- ing these models is to incorporate the chemical heterogeneity by providing a distribution of sites, competition among metals and protons, and a sub-model that describes the activity correc- tions resulting from the charge associated with the metal-ligand complexes (Milne 2000). In most cases, models for metal-humic acid complexation have utilized either discrete site or continu- ous site distributions. Two of the most well accepted models are WHAM Model VI (Windermere humic aqueous acid model) developed by Tipping et al. (1998) and the NICA-Donnan model (Kinniburgh et al. 1996). Both models assume two classes of binding sites attributable to carboxylic and phenolic groups (Tipping et al. 1998; Milne et al. 2001); however, at low metal concentration non-uniform binding and nitrogen and sulfur binding sites may be important (Frenkel et al. 2000). Even though these models incorporate the basic tenants required for describing metal-ligand binding in complex sys- tems, they each apply simplifications such as averaging the size, shape, and functional group behavior of humic substances and congruency of site heterogeneity for different metal ions. For example, each class of sites contains multiple binding constants whose values are distributed around a central log equilibrium constant. Both models have adopted the Donnan approach to account for electrostatic effects. Model VI utilizes a discrete distribution of sites assuming multidentate ligands (e.g., bi- and tridentate) in which humic substances are viewed as rigid spheres of uniform size with functional groups positioned on the surface. A large number of parameters are required to describe the site density, median values for the binding con- stants, and the distribution of constants around the means, but a number of assumptions are used to reduce the number of adjustable parameters. Nevertheless, a large number of parameters are required for the model as shown in Table 3-8. Reasonable fits to a range of data have been observed for this model as well as with the NICA-Donnan model, which assumes that humic substances behave as a gel with a homoge- neous charge and potential distribution. The NICA-Donnan was able to fit metal binding data from 124 data sets encom- passing 23 metal ions over 16 orders of magnitude in metal ion concentrations from pH 2-16 using generic log K values for the two binding sites (Milne et al. 2001; Milne et al. 2003). In a recent head-to-head comparison of two models for predicting Cu and Pb metal ion speciation for a soft water river, a hard water lake, and a hard water stream, it was concluded that predictions of the dominant species for each metal agreed reasonably well between the two models and the experimental data; however, both models under-predicted the concentrations of free metal ion concentrations by several orders of magnitude for the stronger complexing metals, Pb and Cu, when these con- centrations represented only a small fraction of the total metal ion (Unsworth et al. 2006). Van Riemsdijk et al. found similar results for Ni complexation in six surface waters tested with Model VI. The model showed that the fraction of free metal ion in the waters increased with increasing Ni concentration, water hardness, and decreasing pH, and the model predictions overestimated the free Ni concentration even when all of the DOM was assumed to be comprised of the more reactive spe- cies, fulvic acid. Cheng and Allen (2006) demonstrated that the prediction of Zn2+ free metal ion concentrations were reasonable at lower pH (below 7.0), but zinc binding for the natural waters tested was overestimated by Model VI at zinc concentrations below 10-6 M at pH 8.0. Better results were obtained by Bryan et al. (2002) for estimating Cu2+ concentrations in 15 freshwater samples in Parameter Description nA Density of type A sites (mol/g) nB Density of type B sites (mol/g) pKA Intrinsic proton dissociation constant for type A sites pKB Intrinsic proton dissociation constant for type B sites pKA Distribution term to modify pKA pKB Distribution term to modify pKVB Log KMA Intrinsic Equilibrium Constant for Metal Binding to Type A Site Log KMB Intrinsic Equilibrium Constant for Metal Binding to Type B Site LK1 Distribution term to modify log KMA LK2 Distribution term to modify bidentate and tridentate binding strengths P Electrostatic parameter Ksel Selectivity coefficient for counterion accumulation fprB Fraction of proton sites that can form bidentate complexes fprT Fraction of proton sites that can form tridentate complexes M Molecular Weight of Fulvic Acid and Humic Acid r Molecular radius of Fulvic Acid and Humic Acid Table 3-8. A summary of Model VI parameters (Tipping et al. 1998).

26 which both humic and inorganic complexation were included using Model V and Model VI. The modeling effort incorpo- rated competition with Mg, Al, Ca, Fe(II), Fe(III), and Zn, and the calculated Cu2+ concentrations from Model VI were within a factor of 3.6 in 95% of cases. The application of these models to various natural waters has been tested for the past decade, and refinements to the approach to estimating parameters, predicting competition, and application to macrophytic organic matter have been a major focus. For example, Rey-Castro et al. (2009) developed a methodology which utilizes analytical expressions or aver- age affinity distributions for the NICA-Donnan model and David et al. (2010) evaluated competition using the NICA model with a Conditional Affinity Spectra of the H+ binding at fixed total metal concentrations (CAScTM). In the study by Rey et al. (2009), their approach allowed them to distinguish the binding locations of different groups of cations in a multi-component mixture: (a) Al, H, Pb, Hg, and Cr preferentially bound to the phenolic sites of fulvic acid; (b) Ca, Mg, Cd, Fe(II), and Mn preferred the carboxylic sites in the multi-component mixture; and (c) Fe(Ill), Cu, Zn, and Ni occupied both the phenolic and carboxylic sites. Using the Windermere Humic Aq. Model (WHAM) which incor- porates Model VI, Tipping et al. (2008) showed that complex- ation constants for trace metal accumulation in bryophytes were similar to those for humic substances suggesting that these models can also be applied to plant uptake. 3.2.5 Summary Metal ions in solution are always associated with other ions, organic species, or particulate matter. Their speciation is a function of the pH, ionic strength, and presence of other ions, organic molecules, and solid phases. Their speciation often dic- tates the mobility and toxicity in a particular aquatic environ- ment because complexation reactions prevent metal ions from being precipitated, complexing agents act as carriers for trace elements in water, and metal ion bioavailability is often reduced by complexation. Simple inorganic ligands such as carbonate can control speciation in natural waters only in the absence of significant concentrations of stronger organic ligands. Natu- rally occurring and anthropogenic chelating agents that have binding constants that are orders of magnitude higher than the simple inorganic and organic acid functional groups can also dictate speciation if present at sufficient concentration. Because metal-ligand complexes can have significantly differ- ent charge than the metal ion itself, the mobility of these spe- cies can be increased dramatically. For example, Cu-EDTA species are typically negatively charged and highly mobile. As with other complexing agents, humic substances interact with and change the solubility and toxicity of trace elements in natural aqueous environments. While general trends in com- plexation with NOM have been observed, complexation with NOM is more difficult to define due to the complex and ill- defined nature of these organic complexes and the variability in the number and types of functional groups. Recent studies suggest that the reactive components of NOM can be charac- terized using a series of simple organic acids that range from weakly binding carboxylate and phenolate to stronger bind- ing salicylate, citrate, and phthalate molecules. In general, complexation of these functional groups with metal cations increases with pH, decreases with metal ion concentration, and increases with fulvic and humic acid concentration. Chemical speciation models provide accurate tools for pre- dicting chemical speciation in water. Speciation of metal ions with inorganic ligands, organic acids, and well-characterized anthropogenic ligands are available in a variety of chemical equilibrium software programs. Challenges associated with describing complexation with NOM still remain. Several models have emerged over the past several decades, and the two most promising models for characterizing speciation with organic matter are the NICA-Donnan model and Model VI. These models have demonstrated ability to predict metal ion binding over a range of conditions and represent the current state-of-the-art. 3.3 Oxidation/Reduction Processes One of the key factors that controls reactivity and bioavail- abilty of metals is the oxidation state of the metal in the sys- tem. Oxidation state is a number that is determined by the difference between the number of electrons associated with an atom in a compound as compared with the number of elec- trons in an atom of the element. Most of the divalent trace metal cations exhibit little redox activity in natural water col- umns indicating that they are typically present in only one oxidation state. Zn and Cd are examples of trace metals that do not typically undergo oxidation or reduction within natu- ral waters. However, several metals including Cr, Se, Co, Pb, As, Ni, and Cu are redox active in natural systems and can exhibit different levels of toxicity, mobility, and bioavailabil- ity depending on the oxidation state of the metal ion. For example, the two major oxidation states of Cr are Cr(VI) and Cr(III). The most common form of Cr(VI) is as an oxyanion (e.g., CrO4=) which is extremely soluble, mobile, and toxic. Chromic acid is a relatively strong acid with pKa’s of 0.86 and 6.51 for the first and second proton dissociation, respectively. In contrast, Cr(III), the reduced form, is relatively insoluble (e.g., Cr(OH)3(s)). In solution, it is typically present in cat- ionic form, sorbs to many oxide minerals, complexes with natural and anthropogenic ligands, and forms both pure phase and mixed precipitates. Remediation strategies for Cr often involve reduction of Cr(VI) to Cr(III), followed by

27 precipitation of a chromium hydroxide solid (Gillham et al. 1994; Wilkin et al. 2005). In general, trace metals do not dictate the redox conditions of a natural water. Rather, biochemical reactions involving major ions, nutrients, and minerals determine the availability of elec- trons in the water, and hence, the redox potential (oxidation/ reduction status) of an element. Trace metals speciate based on the prevailing redox conditions. It is typical to represent the redox conditions of a water based on the electrical poten- tial of the system, EH (volts), or more typically the pe of the system which is related to the electrical potential by: = ln10 E RT F peH where F is Faraday’s Constant (96484.56 Coulombs/mole), R is the ideal gas constant (8.314 Joules/degree Kelvin), and T is the temperature in degrees Kelvin. Natural waters are classified as oxic, suboxic, and anoxic based on their pe (or EH) and pH values as shown in Figure 3-17. Thus, the key to understand- ing redox speciation in natural waters is to determine which redox couple (e.g., Fe+2/Fe+3, O2/H2O, NH4+/NO3-) controls the pe of the system. If the controlling redox couple is at equi- librium, then that couple defines the pe of the system. This pe can be calculated based on known values of the equilibrium constants for the so-called half-reactions for the two species. For example, the half reaction for O2(g) in equilibrium with H2O(liquid) is 1 4O H 1 2H O l log K 20.752 2g e)( )(+ + ⇔ =− + Thus, in an oxic water zone at 25°C, pH 7, and a partial pres- sure of oxygen (PO2) equal to 0.21 atm, the O2/H2O couple controls the redox environment, and the pe of the system will be equal to 13.5. The speciation of trace metals such as As, Cr, Cu, and Ni can then be calculated using the appropriate half reaction corresponding to a redox potential of 13.5. For example, the half reaction for the couple Cu2+/Cu+ is + ⇔ =+ − +Cu e Cu log K 2.62 and the ratio of Cu2+/Cu+ at a pe of 13.5 (EH = 800 mV) is 1010.9, which suggests that Cu2+ is the stable copper species at neutral pH in an oxic water zone. Controlling redox couples for natural waters and soil systems are typically defined by the nutrient conditions prevalent in the system as shown in Fig- ure 3-17. In oxic zones, reduced species such as H2O, N2, and Mn2+ will be oxidized to their corresponding oxidized forms (e.g., O2, NO3-, and MnO2, respectively). In suboxic and anoxic zones, the oxidized forms are reduced. The particular species formed during reduction depends both on the redox potential of the system (e.g., pe or EH) and the presence of microorgan- isms capable of catalyzing the reduction reaction. As stated earlier, the redox potential in a system is controlled or defined by the most oxidized couple as long as the species are present. For example, if O2 is present in a system, the O2/H2O couple will control the EH of the system. As the oxidized species is reduced and disappears from the system, the next couple in the sequence controls. It is also possible, to examine speciation of metal ions as a function of both pH and pe. pe vs. pH graphs can be prepared for each metal ion, and as long as the pe and pH of the sys- tem are known and equilibrium is established, speciation can be predicted. For example, As(III) and As(V) are both typi- cally found as oxyanions, arsenite and arsenate, respectively. Metals such as, As, Se, and Cr can cause water molecules Figure 3-17. The oxidation (oxic zone) and reduction sequence observed in typical soils systems (adapted from Essington 2005).

28 to deprotonate completely, forming the oxyanions AsO3 3-, SeO3 2-, and CrO4 2-. These basic oxyanions undergo acid/base chemistry, forming either di- or tri-protic acid. Thus, arsenic can exist in solution in two different oxidation states and seven different states of protonation. Figure 3-18 presents the speciation of these seven differ- ent forms of arsenic as a function of pH and pe. The stability region for water is also shown on this graph. The pKa for H3AsO3 of 2.24 is shown as a vertical line in Figure 3-18 that represents the pH at which equal concentrations of H3AsO3 and H2AsO3- are present. The vertical line is terminated at the pe at which the oxidized forms are in equilibrium with the reduced species As(OH)3. Diagrams such as the one presented for the arsenic system can be extremely useful for understand- ing chemical speciation and for developing remediation strate- gies for water. Arsenite is more toxic and less reactive, but can be oxidized to arsenate through homogeneous and heterogeneous (e.g., on mineral surfaces) oxidation processes (Sharma and Sohn 2009). For comparison purposes Table 3-9 shows that the ranges of Eh and pH for various freshwater systems are within ranges that encompass both oxidized and adsorbed arsenic species. The importance of adsorption to mineral surfaces to facilitate heterogeneous oxidation cannot be underestimated because homogeneous oxidation occurs slowly. Thus, understanding the distribution of arsenic species and potential remediation options for contaminated water requires an understanding of redox chemistry, as well as sorption reactions and precipitation reactions. In general, equilibrium speciation is less appropriate for predicting the speciation of redox active metal ions because transformations between redox states are often kinetically con- trolled. Rates of oxidation are controlled by a number of water quality parameters. Craig et al. (2009) used a muti-parametric technique to demonstrate that Fe(II) oxidation rates were dependent on carbonate/bicarbonate, NOM, sulfate, chloride, the sulfate/fluoride interaction, and fluoride, in which the most influential parameters are listed first. -1 -0.5 0 0.5 1 1.5 0 2 4 6 8 10 12 14 Eh (v olt s) pH H3AsO4 H2AsO4- HAsO42- AsO43- As(OH)3 As(OH)4- AsO2OH2- Figure 3-18. Arsenic speciation as a function of pH and redox potential (adapted from Cheng et al. 2009). Type of Waters Specific Conductance (mS/cm) EH (mvolts) pH Rain water 2 – 100 +400 to +600 4 – 7 Freshwater lakes/streams 2 – 100 +300 to +500 6.5 – 8.5 Groundwater 50 – 50,000 -200 to +100 6 – 8.5 Brines Up to 500,000 -200 to -600 Near neutral Ocean water ~ 50,000 +300 to +500 7.8 – 8.4 Wetlands/bogs 50 – 50,000 +100 to -100 variable Table 3-9. Typical ranges of pH and Eh for various water bodies (Sanders 1998).

29 While the importance of redox chemistry is evident for redox active metals, it is important to recognize that while metals such as Zn and Cd are not directly influenced by the redox condi- tions of a system, their fate is dependent on the redox state of elements such as sulfur. In a reducing environment, rich in electrons, sulfate is reduced to sulfide and metal sulfides are extremely insoluble. Hence, in reducing environments many divalent metal ions, including Zn and Cd, can precipitate from solution [e.g., ZnS(s) and CdS(s)]. Changes in redox conditions have been shown to affect speciation in the water column of some lakes. Balistrieri et al. (1994) showed that dissolved Co, Cr, Fe, Mn, Ni, Pb, and Zn concentrations increased across the oxic- suboxic boundary (pe = 7 or EH = 400) whereas dissolved As, Co, Cr, Fe, Mn, and V concentrations increased in the anoxic layer (pe < 2 or EH < 100) of a meromictic lake. At greater depths, in the anoxic zones where sulfide was evident, significant decreases in dissolved Cu, Ni, Pb, and Zn concentrations were observed as well as increases in acid soluble particulate concentrations of As, Cr, Cu, Fe, Mo, Ni, Pb, V, and Zn. 3.3.1 Summary Several metals including Cr, Se, Co, Pb, As, Ni, and Cu are redox active in natural systems and can exhibit different lev- els of toxicity, mobility, and bioavailability depending on the oxidation state of the metal ion. The particular oxidation state of a metal is typically controlled by dominant biogeochemical processes occurring within the system. Several metals such as Cr and As exhibit multiple oxidation states in natural waters and the relative rates of oxidation/reduction are often controlled by interactions with NOM and surfaces. In addition, metal fate and transport is affected by the dominant redox processes that affect speciation of major anions such as nitrate and sulfate. In reducing environments, nitrate reduction to ammonia increases, metal ion complexation, and sulfate reduction to sul- fide can lead to precipitation of many of the trace metal ions. 3.4 Precipitation Processes Precipitation processes are also typically rate limited, but may control the fate and transport of metal ions in natural waters due to the formation of mixed precipitates and meta- stable phases. Precipitation reactions are typically defined by Ksp or solubility reactions written in terms of the dissolution of the ionic solid. For example, CuCO s Cu CO Ksp 10 CuOH s Cu 2OH Ksp 10 3 2 3 2 9.63 2 2 19.36 ) ) ( ( ⇔ + = ⇔ + = + − − + − − The extent and rates of precipitation or dissolution are dependent on the solution phase concentrations of both spe- cies as well as the pH of the system. In all cases, precipitation occurs via a series of steps including nucleation, growth, and crystallization. The solution pH determines both the OH- concentration and the concentration of other reactive ligands such as CO3 2- , which undergoes acid/base reactions to main- tain equilibria with HCO3- and H2CO3. Rates of precipitation are dependent on the degree of oversaturation defined by: IAP Ksp Cu CO Ksp 2 3 2{ }{ } = + − where IAP is the ion activity product determined from the product of the measured activities (NYSDEC) of the reaction products divided by the product of the activities of the reac- tants relative. If the ratio of IAP/Ksp is greater than one, then the system is oversaturated with respect to the solid phase. The rate of precipitation as a function of the relative super- saturation is typically expressed as a power law ( )( )= −R k IAP Ksp 11/2 n where k is the apparent rate constant and n the apparent rate order. The degree of oversaturation often determines whether homogeneous or heterogeneous nucleation dominates. In het- erogenous precipitation, a different surface or substrate acts as the site for ions of the crystal to properly orient whereas in homogeneous nucleation a few particles correctly orient in the course of their random movement through the water. One of the systems most commonly studied with respect to rates of precipitation, homogeneous and heterogeneous nucleation is the CaCO3(s) system. Numerous studies have investigated the relative degree of oversaturation required for homogeneous nucleation and the potential of other natural water particles to facilitate heterogeneous nucleation. Lin and Singer (2005) showed that at low degrees of supersaturation a nucleation site from another surface is required to overcome the activation energy barrier for precipitation. In their study of quartz, dolo- mite, and commercially available calcite and lime softening pre- cipitates, only the calcite-based seed materials were capable of inducing calcite precipitation within a 2 hour time period and required a degree of oversaturation of greater than five. Impu- rities within the seed such as Mg2+ and NOM were reported to have poisoned the seed and reduced its catalytic ability. Other researchers have demonstrated similar effects at low supersatu- ration ratios. Lioliou et al. (2007) demonstrated that induction rates for homogeneous nucleation were proportional to the degree of supersaturation and suggested a polynuclear growth mechanism. At lower supersaturation ratios, the addition of calcite seed crystals led to immediate precipitation without an induction period. However, addition of quartz crystals did not lead to precipitation. Again the differences were attributed to

30 the lack of compatibility between the crystal lattices. Both of these studies emphasize the importance of lattice compatibil- ity for nucleation. Similar findings have been reported for co- precipitation processes in which ion size is a key determinant for removal of trace metal ions in precipitating iron, alumi- num, and silica systems (Maurice 2009). It is important to also recognize that precipitation of a meta- stable phase may often precede formation of the most thermo- dynamically stable phase for a specific metal. For example, the presence of high concentrations of iron in water often leads to precipitation of amorphous iron hydroxide [Fe(OH)3] in rela- tively short time periods (hours) whereas it takes significantly longer times to form stable goethite (FeOOH) or hematite (Fe2O3) phases. As each of these more stable phases is formed, the precipitate becomes more crystalline with a decreased SSA as shown in Figure 3-19. The observed relationship between the crystallinity of a min- eral and the SSA is typical of many oxide and hydroxide miner- als (Schwertmann 1996). For iron oxides, the SSA is greatest for ferrihydrite and then decreases as the amorphous structure ages to the more crystalline structures of goethite and hematite. As a result, it is often necessary to calculate chemical speciation based on the operative redox state or based on knowledge of the metastable solid phase present. The precipitation of metal hydroxides, oxides, carbon- ates, and sulfides can control the fate and transport of metal ions. Thermodynamic data for precipitation of metal species are tabulated in databases such as the Critical Stability Con- stants (Martell and Smith 1974), geochemical speciation data- bases contained in models such as MINTEQA2, MINEQL+, PHREEQC, and others. In many cases, values obtained from these sources differ significantly (by orders of magnitude in some cases) due to differences in experimental methods, parti- cle size and crystallinity, experimental temperatures, pH values, aging times, analytical limitations, etc. Dyer et al. (1998) reviewed solubility constants for a num- ber of hydroxide solids and developed speciation diagrams for the more amorphous phases documented in the literature. An example of the type of solubility diagram developed is shown in Figure 3-20 for two metal hydroxide solids, Cu(OH)2(s) and Zn(OH)2(s). The solid line in these figures represents the equi- librium between the solid precipitate and the total aqueous con- centration of the metal. The curvature in the solubility line is due to complexation reactions which change the dominant sol- uble species [e.g., Cu2+ to Cu(OH)+ to Cu(OH)2 0 to Cu(OH)3-] as described previously. Thus, the line represents the maximum solubility of the metal ion as a function of the pH of the water. Waters having concentrations of metal ion that fall above the line are oversaturated with respect to the metal ion. Dyer et al. (1998) compiled model simulations for 12 dif- ferent metals and their total solubilities in aqueous solution. Their data highlighted two important points. Fe(OH)3(s) and Al(OH)3(s) have minimum solubilities at lower concentrations and lower pH than all the other metal hydroxides except Ni(II). In addition, most of the trace metal ions have minimum solu- bilities that are above 0.1 ppm in the pH 6 to 8 range, suggest- ing that hydroxide precipitates will not control the mobility of trace metal contaminants in most systems. These results sug- gest that iron and aluminum hydroxides represent important sinks for metal ions and explain why these solids are frequently used in water treatment processes for coagulation processes. Zn, Ni, and Co have been studied for their potential to sorb to aluminum oxides and clay minerals in laboratory and field studies. A thorough discussion of the adsorption process is pre- sented in Chapter 5. However, not only do these metals adsorb to the surfaces of clays and iron, silica, and aluminum oxides and hydroxides, they also co-precipitate with them. The for- mation of mixed-metal-Al layered double hydroxide (LDH) phases similar to hydrotalcite has been well documented in the literature. The rates and thermodynamics of formation of these phases depend on a number of scenarios including the nature of the aluminum phase and the type of interlayer anion pres- ence (Scheidegger et al. 1998; Thompson et al. 2000). Substitu- tion of silica for carbonate and carbonate for sulfate led to a Amorphous Crystalline 100-400 m2/g 8-200 m2/g 2-90 m2/g Ferrihydrite, Fe5HO8·H2O Goethite, -FeOOH Hematite, -Fe2O3 Figure 3-19. SA as a function of crystallinity for select iron oxides (adapted from Cornell and Schwertmann 1996). Figure 3-20. Comparison of Cu(OH)2 and Zn(OH)2 solubility model predictions.

31 more stable phase (Peltier et al. 2006). Again, the key to metal ion substitution is similarity in ionic size between the parent ion and the substituting cation. Thus, the presence of clay par- ticles that can provide both aluminum and silica via dissolution can promote the formation of these mixed precipitates. Similar types of mixed metal precipitates have been identified for nickel and cobalt silicates in which the silica is derived from silica minerals. In all cases, the presence of these phases lowers the overall solubility of the metal ion. Thus, the solubility of both the trace contaminant phases and the dominant base mineral phases in the system will dictate the overall solubility of the metal contaminants. For example, thermodynamic data suggests that Zn hydroxide [Zn(OH)2], smithsonite (ZnCO3) or hydrozincite [Zn5(OH)6(CO3)2] will control Zn solubility in contaminated neutral and calcareous soils (Saeed and Fox 1977; Jacquat et al. 2008). Jacquat et al. (2008) studied zinc contami- nated soils and identified the presence of Zn-phyllosilicate and Zn-LDH phases from field samples. Not only did the pre- cipitation of these phases control the total zinc concentra- tion in solution at low and intermediate surface loadings of Zn, but separate experiments with synthetic phases suggested that zinc could be readily extracted from these phases using 1 M NH4NO3 followed by 1 M NH4-acetate at pH 6.0. The implications of these studies are that the presence of domi- nant ions such as iron, aluminum, and silica in natural waters may impact trace metal concentrations through precipitation reactions even when the metal ions are present at levels below their saturation. 3.4.1 Summary Precipitation processes can have a large impact on the speciation of metal ions in solution. In oxic waters, the major precipitates include hydroxide and carbonate phases. While published solubility constants are available for most of the potential solid phases, significant variability exists in the litera- ture due to the experimental methods and conditions used to determine the values. Moreover, precipitation processes are often rate limited, and the most stable phase is often not the dominant phase found in natural waters. The kinetics of pre- cipitate formation are dependent on the presence of nucleation sites from other similar phases or from homogenous solid for- mation. The evidence is also mounting that removal of metal ions in natural and engineered systems is due to the more rapid formation of mixed precipitates or co-precipitates formed from either dissolution of prevailing solid phases (e.g., clays and oxide minerals) or precipitation of more insoluble phases such as iron and aluminum hydroxide. Removal of metal ions from solution may also be controlled by adsorption to these other precipitating phases as is commonly noted in water treatment applications. 3.5 Sorption Processes One of the dominant mechanisms for removing metal ions from natural waters is sorption to suspended solids, colloidal solids, and other particulate matter present in the system. By definition, sorption involves either adsorption or absorption, although macroscopic measurements of sorption rarely distin- guish precipitation and surface precipitation from these other two processes. The term adsorption typically refers to the accu- mulation of compounds on a solid surface, and absorption is related to accumulation in the internal region of a solid phase. Inorganic species, such as metal cations, have the potential to partition to solid surfaces present in aquatic systems. This interaction, generally termed sorption, involves distributing the metal species between the solid and aqueous phases in a man- ner analogous to the complexation reactions that occur in solu- tion. Indeed, the interaction at the solid surface often involves the reaction between the metal species with surface functional groups, such as surface oxygens and hydroxyl groups. The extent of metal ion sorption is largely dependent on the type of solid mineral present, bulk aqueous concentration of the metal ion, solid concentration, pH, ionic strength, and the presence of aqueous ligands. The properties of solid minerals present in nat- ural systems vary significantly for different minerals and surface functional groups; therefore, it is important to understand solid phase characteristics when discussing metal ion adsorption. 3.5.1 Charge Development in Clays and Clay Minerals Both clays and clay minerals such as oxides, hydroxides, and oxyhydroxides provide surfaces for adsorption of metal ions. All of these minerals create charged surfaces when placed in solutions. Clays are comprised of sheets of silica tetrahedral and aluminum octahedra that are stacked as either 1:1 or 2:1 layers as shown in Figure 3-21, with the dimensions shown in angstroms (Ao). Two types of charge are possible on the sur- faces of clays: permanent charge surfaces and variable charge surfaces. Permanent charge on clays develops from isomorphic substitution of aluminum for silica in the tetrahedral sheet and Mg and other divalent metal ions for Al in the octahedral sheets. The substitution of an ion with lower charge [i.e., Al(III) for Si(IV) and Mg(II) for Al(III)] yields a net negative charge on the inner layer of the clay mineral as shown for montmorillonite in Figure 3-21. For expanding clays such as montmorillonite, this permanent charge is balanced by inner layer cations such as Ca2+ and Mg2+. Ion exchange among these inner layer cat- ions provides one mechanism for removal of metal ions from solution. However, since trace metals are present at substantially lower concentrations compared to the major cations of the sys- tem, competition with major cations must be considered. In addition to the charge that develops due to isomorphic substitution, clays also develop charge on their edges where

32 where ≡S represents a generalized surface site. The pH at which the concentration of [≡SO-] is equal to the concentration of [SOH2+], is termed the pH of the point of zero charge (pHpzc). At this pH, the net charge on the surface is neutral. Below the pHpzc, the surface will carry a net positive charge and above the pHpzc, the mineral will carry a net negative charge. Clay minerals such as oxides, hydroxides, and oxyhydrox- ides also contain terminal edge sites which develop charge in solution. For example, the goethite surface (FeOOH) is comprised of oxygen-hydroxyl sheets that form a hexagonal close-packed array in which iron resides in two-thirds of the oxygen octahedra, but in strips as shown in Figure 3-23. In the goethite structure, surface oxygens can be singly, dou- bly, or triply coordinated to iron atoms, providing a range of reactivities. However, it is difficult to estimate differences in the reactivities of these different groups, and many researchers Figure 3-21. Structures of kaolinite, a 1:1 clay (top) and montmorillonite, a 2:1 clay (bottom). The arrow in the structure for kaolinite indicates the location of hydrogen bonding which gives kaolinite its rigid structure and prevents inner layer ion exchange. Figure 3-22. Terminal bonds on the edge of a 1:1 clay. oxygens terminate (Figure 3-22). These terminal oxygens of the clay are not fully coordinated as in the bulk clay struc- ture and tend to protonate in water. As a result, they are often characterized as diprotic surface sites which undergo the following acid/base reactions: SOH SOH H Ka SOH SO H Ka 2 1 2 ≡ ⇔ ≡ + ≡ ⇔ ≡ + + + − + Figure 3-23. Goethite structure displaying octahedra formed by oxygen atoms.

33 use a single set of diprotic acidity constants to describe proton release from these surfaces. This approach has been questioned as of late, and significant progress has been made that dem- onstrates that differentiating adsorption to different site types yields significant improvements in model fits to metal ion adsorption to geothites (Villalobos and Perez-Gallegos 2008; Villalobos et al. 2009; Salazar-Camacho and Villalobos 2010). Of particular interest with respect to the acidity of these sites are the differences among the different oxide minerals as shown by the variation in pHpzc values in Table 3-10. Since above the pHpzc, the surface carries a net negative charge, min- erals with a lower pHpzc exhibit a higher tendency to attract cations. In contrast, minerals with a higher pHpzc have greater affinity for anions. Surfaces other than oxides also provide reactive sites for binding metal ions. Carbonate minerals also affect the bio- geochemistry of natural waters and create surfaces for adsorp- tion. Reactive sites on these minerals include ≡CaOH and ≡CO3H. On this surface, Ca, in addition to protons, affects the intrinsic surface charge as surface reactions including: CO H CO H CO H CO H H CO H Ca CO Ca H CaOH CaO H CaOH H CO CaHCO CaOH CO CaCO 3 3 3 2 3 3 2 3 3 2 3 0 3 2 3 ≡ ⇔ ≡ + ≡ ⇔ ≡ + ≡ + ⇔ ≡ + ≡ ⇔ ≡ + ≡ + + ⇔ ≡ ≡ + ⇔ ≡ − + + + + + + − + + − − − Thus, the surface charge characteristics of calcium carbonate surfaces is complex. At low pH, adsorption of bicarbonate and calcium ions determine the surface charge. As pH increases, the adsorption and desorption of protons becomes the dominant source of charge. In addition, many complications arise in determining the surface charge of calcite and other carbonate minerals due to the experimental challenges associated with establishing and maintaining equilibrium with the atmosphere which contains carbon dioxide and with the solid phase (Wolthers et al. 2008). Dissolution of carbon dioxide and precipitation of calcium car- bonate are both pH dependent properties. The total aqueous carbonate concentration and the carbonate concentration both increase dramatically with increasing pH, and the time frame for establishing equilibrium with the atmosphere is on the order of hours and with the solid phase can be even longer. Thus, many experimental studies examining surface charge and adsorption to carbonate surfaces have been impacted by non- equilibrium artifacts. Thus, it is not surprising that the pHpzc for calcite has been reported over a range of values from pH 7 to 11. 3.5.2 Surface Complexation One of the most intriguing aspects of metal cation sorption is the ability of some metal ions to sorb at pH values below the pHpzc. The sorption reaction, termed surface complex- ation, is thought to be analogous to the aqueous complex- ation of metal ions by ligands present in the bulk solution. The formation of surface complexes by divalent metal cations is often described by the following general chemical reaction: ≡ + ⇔ ≡ ++ +SOH Me SOMe H2 In this reaction, ≡S represents the generalized surface site acting as a complexing ligand. As a result of surface com- plexation, metal ions can sorb to oxide surfaces at pH values below the pHpzc as shown for adsorption of Co by aluminum oxide in Figure 3-24. These pH adsorption edges are obtained from batch experiments in which identical concentrations of metal ion and solid (oxide mineral) are added to a series of reactors and the pH is varied through addition of strong acid or base. The shape of the adsorption edge is consistent with the complexation reaction shown above and suggests that for strongly sorbing metal ions such as cobalt, there is an increase in adsorption from near zero to 100% removal from solu- tion over a fairly narrow pH range. The data also show that increasing the solids concentration or decreasing the amount of Co added to the solution shifts the edge to the left, which indicates increasing adsorption. More typical isotherm plots may also be generated from metal ion adsorption data in which the solid phase concentra- tion (or mass adsorbed/mass of sorbent) is plotted versus the aqueous phase equilibrium metal ion concentration as shown in Figure 3-25 for adsorption of Cu on iron oxide between pH 6 and 7. These model simulations (using the Diffuse Layer δ-SiO ) Mineral pHpzc Quartz (α-SiO2) 2.9 Amorphous silica (SiO Birnessite ( 2 2H2O) 3.5 2 3.76 Kaolinite 3.76 Rutile (TiO2) 5.8 Anatase (TiO2) 6.0 Magnetite (Fe3O4) 6.9 Muscovite 7.5 -Alumina (γγ -Al2O3) 8.5 Hematite (α-Fe2O3) 8.5 Gibbsite (Al(OH)3) 8.9 Corundum (α-Al2O3) 8.9 Goethite (α-FeOOH) 9.0 Table 3-10. Point of zero charge values for various clays and clay minerals (Sverjensky and Sahai, 1996).

34 Model – DLM) show the non-linearity of the isotherms and the importance of incorporating pH into modeling approaches. As a result, descriptions of metal ion adsorption that utilize linear isotherms in which the sorbed phase concentration is related to the equilibrium solution phase concentration by a distribution coefficient, Kd, are inadequate for describing metal ion adsorp- tion. In addition, non-linear empirical models such as the Langmuir and Freundlich isotherm models are also inaccurate, as they do not adequately capture the effect of pH. Figure 3-26 shows the dramatic difference in adsorption of weakly sorbing Se(IV) and strongly sorbing Se(VI). The data also show a large impact of ionic strength on sorption of Se(IV) with reduced sorption at higher ionic strengths. The inability to achieve 100% adsorption at low pH in these sys- tems can be due to site limitations or the change in speciation from the fully deprotonated oxyanion to a protonated acid form with lower charge (i.e., SeO4 + 2H+ ⇔ H2SeO4). Similar results have been reported from other oxides and other oxy- anions (Essington 2005). The presence of multiple adsorbing cations or cations and anions can lead to competition for sites on the surface. Competition for surface sites is just one possible affect that has been observed in multi-solute systems. The adsorption of oxyanions can reduce the negative surface charge on oxide minerals thereby promoting adsorption of cationic metals (Boyle-Wight et al. 2002; Zhang and Peak 2007). In addition, evidence for ternary complexes (see Figure 3-27) has been presented by a number of researchers and its impact on adsorption (Zhang and Peak 2007; Papadas et al. 2009; Swedlund et al. 2009). Thus, it appears that both competi- tion and enhanced adsorption are possible in the presence of sorbing ligands. 0.2g/L 20 g/L 2.0 g/L Co=2x10-6M 0 20 40 60 80 100 8765 pH % A ds or be d . 9865 Co=1x10-4 M Co= 2 x10-6 M l 20g/L 7 Figure 3-24. Sorption of Co(II) to a-Al2O3 (aluminum oxide, corundum) at constant initial Co concentration and aluminum oxide concentration as a) a function of aluminum oxide concentration and b) a function of initial Co concentration. 10 9 8 7 6 5 4 3 2 1 0 14 12 10 8 6 4 2 0 lo g So rb ed Ph as e Co nc en tr ati on (m ol /g ) log Aqueous Phase Concentration (mol/L) pH 7 DLM Simulation pH 6.4 DLM Simulation pH 6 DLM Simulation Cu(II) Adsorption Isotherm Simulations Hydrous Ferric Oxide Figure 3-25. Effect of pH on sorption isotherms for Cu onto iron oxide (hydrous ferric oxide). Model simulations generated using Visual MINTEQ.

35 Organic acids also adsorb to mineral surfaces. A number of researchers have studied the adsorption of organic acids and shown that they form inner sphere complexes that can affect both adsorption and dissolution rates of minerals (Ali and Dzombak 1996; Hwang et al. 2007; Hwang and Lenhart 2008a; Hwang and Lenhart 2008b; Hwang and Lenhart 2009; Hwang and Lenhart 2010). Research conducted in the mid- to late-1990s concluded that organic acids adsorption is similar to oxyanion adsorption and that carboxylic groups exhibit an adsorption maximum at low pH and phenolic groups have their adsorp- tion maximum at higher pH (Gu et al. 1994; Gu et al. 1995; Evanko and Dzombak 1998). Figure 3-28 shows the proposed structure of two organic acids on the surface of hematite. The adsorption of organic acids can also influence the sur- face charge and hence, the subsequent adsorption of metal cat- ions (Johnson et al. 2005). The structure of the sorbed complex depends on the affinity of the metal and oxyanion for the sur- faces. For example, Buerge-Weirich et al. (2004) showed that the impact of organic acids on Cu, Ni, and Cd adsorption to goethite depended on pH, metal ion, and ligand. Competition for sorption sites led to a decrease in Cu and Ni adsorption at high pH values in the presence of the organic ligand, oxalate. However, below pH 6 surface complexation modeling of Cu adsorption was best described by defining a ternary complex of Type A (surface-metal-ligand). The adsorption of Cu in the presence of salicylate above pH 5 was also to a ternary complex of Type A. Increased adsorption of Cu and Cd in the presence of pyromellitate at acidic pHs was attributed to formation of Type B (surface-ligand-metal) ternary surface complexes. While these results were not verified by spectroscopic methods, they do provide evidence for the importance of both competing cat- ions and complexing ligands for predicting metal ion adsorp- tion to oxide surfaces. Of particular importance in this work is that the ligands selected for study are representative organic ligands used to simulate binding by humic acid moieties. Thus, this data also support the contention that humic and fulvic acids adsorb to oxide minerals. The extension of this work to fulvic acids has been con- ducted by a number of researchers. Figure 3-29 shows the adsorption edges for removal of fulvic acid onto goethite. The adsorption behavior is typical of what is expected for oxyanion 0 10 20 30 40 50 60 70 80 90 100 4.00 5.00 6.00 7.00 8.00 9.00 10.00 pH Pe rc en t S el en iu m S or be d Se(VI) = 0.6mM, I = 0.1 M Se(VI) = 0.7mM, I = 1.0 M Se(IV) = 0.8 mM, I = 0.1 M Al2O3 = 5 g/L Figure 3-26. Adsorption of the oxyanions selenate Se(VI) and selenite Se(V) onto aluminum oxide. I denotes the ionic strength of the solution, M is molarity (mol/L), and mM is mmol/L. Metal like sorption Ligand like sorption Metal and ligand sorption SOH + Me2+ + SeO32-<=> SOMeSeO3 -+ H+ (sorption to surface hydroxyl site through the metal ) M OH H O Se O O SOH + Me2+ + SeO32-<=> SOMeSeO2 -+ H+ (sorption to surface hydroxyl site through the ligand) Se O O M O H H O H H O H H 2SOH + Me2+ + SeO32-<=> 2SOMeSeO2 -+ H+Se O O M O H H (sorption to surface hydroxyl sites through metal and ligand) Figure 3-27. Representative ternary adsorption complexes for a representative oxyanion (SeO3 2) and a sorbing metal cation. Figure 3-28. Proposed structures of adsorbed maleate and fumarate on iron oxide (adapted from Hwang and Lenhart 2008a). Figure 3-29. Adsorption of fulvic acid on goethite (adapted from Filius et al. 2000).

36 and organic acid adsorption; removal decreases with increas- ing pH and increased fulvic acid shifts the adsorption edge to lower pH (Filius et al. 2000). Bäckström et al. (2003) showed that 20 ppm of fulvic acid added to a goethite system followed adsorption trends consistent with anions in which the percent removed from solution decreased from over 90% to less than 10% as the pH increased from 7.5 to 10. Mercury and Cd(II) displayed typical cationic behavior in which adsorption increased with increasing pH. However, in studies of mercury adsorption, the presence of fulvic acid increased adsorption and had the largest impact at low pH. Fulvic acid also increased adsorption of cadmium, but only below pH 7. Above pH 7, the presence of fulvic acid decreased adsorption. Similar results for enhanced adsorption at low pH in the presence of fulvic acid were shown by Wu et al. (2003) for adsorption of Cu and Pb on gamma-aluminum oxide, and Weng et al. (2008) for Cu to goethite in the presence of fulvic acid. These results suggest that pH, ionic strength, complexation with organic acids and fulvic acids, and competition by other adsorbing species must be considered in order to predict free metal ion concentrations in the presence of oxide minerals. Similar evidence for surface complexation to other particu- late matter including the edges of clay minerals, bacteria, and algal matter has also been demonstrated in the literature. 3.5.3 Surface Complexation Models Surface complexation models (SCMs) provide a thermo- dynamically based approach for describing adsorption in which solutes interact with functional groups on the surface either through ion pair association or solute-functional group complexation. In addition, these models also incorporate a description of the electrical double layer (EDL) surrounding the particle that describes the relationship between surface charge and potential, and this description is used to modify the activity coefficients for each surface species (Davis and Leckie 1979; Westall and Hohl 1980; Goldberg 1992). Thus, there are four basic tenants that all SCMs are founded upon: (1) the adsorption of ions occurs at specific sites on the min- eral surface; (2) adsorption reactions can be explained thermo- dynamically through the use of mass-law expressions; (3) the electrostatic effects associated with adsorption are considered; and (4) adsorption of ions results in the surface being charged (Davis and Kent 1990; Dzombak and Morel 1990; Christl and Kretzschmar 1999). A number of SCMs exist today including the Diffuse Dou- ble Layer Model (Dzombak and Morel 1990), the constant capacitance model (Schindler and Kramer 1968; Stumm et al. 1976; Stumm et al. 1980), the triple layer model, the modified triple layer model (Hayes and Leckie 1986; Hayes et al. 1988), and the CD-MUSIC model (Hiemstra et al. 1989a; Hiemstra et al. 1989b; Hiemstra and Van Riemsdijk 1996; Hiemstra et al. 1996). While these SCMs are all based on the four prin- ciples laid out above, each of them describes the structure of the mineral–water interface differently, resulting in mass-law expressions for surface reactions with electrostatic correction terms that are unique to a given model (Davis and Kent 1990). Parameters common to all SCMs include solids concentration (Cs) expressed in g/L, SSA normally in units of m2/g, and sur- face site density (Ns) typically expressed in sites/nm2. The most simplistic description of the EDL incorporated into SCMs is the DLM (see Figure 3-30). In large part, model complexity is related to the complexity of the description of the EDL. Indeed, calculation of the surface potential for the DLM only requires knowledge of the surface charge and the ionic strength of the bulk solution. The constant capacitance model represents only slightly more complexity and is appli- cable to high and constant ionic strength systems. The model shown schematically in Figure 3-31 treats the double layer region as a parallel plate capacitor in which sorbing ions affect the intrinsic charge on the surface and the surface charged decays linearly with charge and distance. The relationship between surface charge and potential is described by a single parameter, the capacitance. In order to more accurately capture the behavior of metal ion sorption over a broad range of ionic strengths, the triple layer model and modified triple layer model were developed. This model incorporates both aspects of the constant capaci- tance model and the DLM as seen in Figure 3-32. Sorbing anions and cations are located on the surface (or o-plane), on O O O Na+ Cl- H H d d O + - - Cl- Charge ( ) =0.1174I0.5sinh(zF d/2RT) Potential ( ) = H2O = Fe Iron Oxide Surface Figure 3-30. Representation of the diffuse layer for the DLM version of the surface complexation model in which d represents the potential at the diffuse layer, sd is the diffuse layer charge, R is the ideal gas law constant, F is Faraday’s constant, z is the ion charge, I is the ionic strength, and T is the temperature in K.

37 a plane located further from the surface (the b plane), or within the diffuse layer depending on the affinity of the sorbing ion for the surface. Two parallel plate capacitors and the DLM are placed in series to determine the potential on the surface as a function of charge. Thus, the model requires estimation of two different capacitances. Finally, the most rigorous model to be employed for surface complexation is the CD-MUSIC model. This model also utilizes the three plane approach but assumes that the surface is heterogeneous. Thus, a different represen- tation of the three plane model is required for each different crystal face on the mineral. The complete description of the surface complexation model also includes reactions for surface protonation and metal ion adsorption to the surface sites, and a value for the surface site density (i.e., the number of sites/m2 of surface). While many trace metal cations bond directly to the surface oxygens in what is termed inner sphere adsorption, weaker binding background electrolytes are assumed to occupy the EDL. Their affinity for the surface is accounted for by the EDL equation shown in Fig- ure 3-30, which incorporates ionic strength. More complicated SCMs utilize more sophisticated descriptions of the double layer with multiple adsorption planes at which strongly and weakly bound cations adsorb, and they include specific reac- tions for binding of background electrolyte ions such as Na+ and Cl-. However, as the models become more complicated, the number of fitting parameters increases. SCMs must be capable of predicting adsorption as a func- tion of solute and sorbent type, solute and sorbent concentra- tion, pH, ionic strength, and the presence of aqueous ligands. Typical model results from the triple layer version of the sur- face complexation model show the effect of solute and sorbent concentration on cation sorption (Figure 3-33). Studies of metal ion sorption to ferrihydrite indicate that sorption will occur in the following order, Pb(II)>Cu(II)> Zn(II)>Cd(II), with increasing pH as depicted in the figure. This behavior correlates with the first hydrolysis constant of a metal cation, the relative acidity of the solid surface, the surface site concentration, and the type of complex formed (Davis and Kent 1990). The hydrolysis constant is directly related to cation acidity; therefore a higher hydrolysis con- stant equates to higher acidity and sorption at lower pH val- ues. Lead, which has its first hydrolysis constant of 106.3, sorbs at a lower pH than Cd, with a hydrolysis constant of 103.9. An increase in the overall initial solute concentration, as shown in Figure 3-33b, results in a flattening of the adsorp- tion edge. An increase in the sorbent concentration, or an increase in the available surface sites, results in the sorption edge shifting to lower pH values, as illustrated in Figure 3-33c. In contrast to the metal cation adsorption edges, oxyanion or ligand adsorption follows a trend in which maximum adsorption occurs at low pH as shown in Figure 3-26. Ionic strength has significant impact on weakly sorbing metal ions Figure 3-31. Description of the EDL for the constant capacitance model (Vieira 2006) where o represents the potential at the surface, d represents the potential at the diffuse layer, s0 is the diffuse layer charge, C1 is the capacitance. Figure 3-32. Representation of the EDL for the triple layer model where (d represents the potential at the diffuse layer, sd is the diffuse layer charge, R is the ideal gas law constant, F is Faraday’s constant, z is the ion charge, I is the ionic strength, and T is the temperature in K (Vieira 2006).

38 The sorption of cations may also be strongly dependent on the presence of competing ions and ligands that can either compete for adsorption sites, or alter the speciation in solution through the formation of non-sorbing aqueous complexes. Stokes (2009) showed that competition between Cd and Zn and Cu and Pb on iron oxide could be modeled using the diffuse layer SCMs; however, the extent of competition was relatively low when both metal ions were added at similar concentrations. Recent work has demonstrated that modeling competitive effects are highly dependent on the surface site density esti- mates. It was found that crystallographic estimates of the sur- face site density, along with an understanding of the reactivity of various sites, enhanced model predictions for all modeling approaches. While the most rigorous modeling approach, CD-MUSIC, seemed to best describe multi-solute adsorption behavior of Pb2+, Cd2+, and SeO3 2- on goethite (Figure 3-35), all of the models benefit from enhanced approaches toward site density estimation. All four of the SCMs described above have been widely used in laboratory studies to describe sorption in increas- ingly complex systems and all of the models are capable of describing single solute sorption in well-defined waters. How- 0% 20% 40% 60% 80% 100% Pe rc en t A ds or be d Pb Cu Zn Cd 0% 20% 40% 60% 80% Pe rc en t A ds or be d Increasing Initial Solute Concentration 0% 20% 40% 60% 80% 2 3 4 5 6 7 8 pH Pe rc en t A ds or be d Increasing Solids Concentration a b c Figure 3-33. Surface complexation predictions of the location of cation sorption edges for (a) varying metal ions, (b) increasing solute concentration, and (c) increasing solid concentration. Figure 3-34. Triple layer model results for outer- sphere sorption of strontium (a) and inner sphere sorption of cobalt (b) on aluminum oxide as a function of ionic strength and solid concentration (for cobalt). (a) 6 Pe rc en t A ds or be d Pe rc en t A ds or be d 87 NaNO3 = 0.01M NaNO3 = 0.5M 109 Sr = 5 x 10-5M α-Al2O3 = 20g/L 6 pH 11 987 1110 Sr = 5 x 10-5M α-Al2O3 = 20g/L NaNO3 = 0.003M NaNO3 = 0.1M 0 20 40 60 80 100 (b) such as Sr, but little impact on strongly sorbing cations such as Co. Such dramatic changes in the adsorption behavior as a function of ionic strength have been used to distinguish inner and outer-sphere adsorption. Model results for triple layer modeling of Co and Sr adsorp- tion to aluminum oxide are shown in Figure 3-34. While Sr and Co are not typical metals found in natural waters, Sr is repre- sentative of other divalent alkaline earth metals such as Mg and Ca and Co is typical of many of the divalent heavy metals such as Cu, Pb, and Zn. Moreover, these results suggest that adsorption of the weakly sorbing metal ions such as Mg and Ca will be significantly impacted by the ionic strength of the water whereas adsorption of heavy metals will be independent of ionic strength.

39 0% 20% 40% 60% 80% 100% 120% 0 50 100 150 200 250 300 350 400 450 2 3 4 5 6 7 8 9 10 11 12 % S e (a q) Io n Ad so rb ed (n m ol /m 2 ) pH Experimental Pb Model Total (Pb) Experimental Se Model Total (Se) H2SeO3 HSeO3[-] SeO3[-2] PbSeO3 Pb(OH) 2 precipitates [Pb] TOT =10 4.07 M [Se] TOT =10 4.02 M [Goethite] = 3.94 g/L Figure 3-35. CD-MUSIC model predictions for Pb(II) and selenite bi-solute adsorption onto goethite (dashed lines represent aqueous species distribution). ever, as the complexity of the system increases with the addi- tion of multiple sorbents, multiple sorbing ions, and humic substances, the models appear less robust. Models such as the extended triple layer model (Sahai and Sverjensky 1997; Criscenti and Sverjensky 1999; Sverjensky 2006; Sverjensky and Fukushi 2006) and the CD-MUSIC model (Hiemstra 1996; Rietra et al. 2001; van and Hiemstra 2006; Weng et al. 2008) show the best potential for accurately predicting sorption in complex systems. However, these models require complete quantification of the solution and solid phase composition, even background electrolyte ions. As a result, several researchers have taken a different approach to modeling metal ion sorption. Instead of increasing the complexity of the model to describe a more complex system, they have used a simplified version of the surface complexation modeling approach in which the EDL properties are ignored. The non-electrostatic model has been shown to provide rea- sonable predictions of sorption in a range of natural systems (Davis et al. 1998; Serrano et al. 2009). While the models do not accurately predict surface charge behavior, their ability to capture sorption over a range of conditions is often sufficient for field applications. The DLM also represents a relatively simplistic model that can be applied readily to both simple and complex sys- tems (Wen et al. 1998; Benyahya and Garnier 1999; Yang and Davis 1999; Lee and Davis 2000; Pretorius and Linder 2001; Swedlund and Webster 2001; Swedlund et al. 2003; Tonkin et al. 2003; Zhou et al. 2005; Bibby and Webster-Brown 2006; Hizal and Apak 2006; Allan et al. 2007; Carroll et al. 2008; Goldberg and Criscenti 2008; Lund et al. 2008; Song et al. 2008; Unuabonah et al. 2008; Hiemstra and Van 2009; Landry et al. 2009; Schaller et al. 2009; Stokes 2009; Swedlund et al. 2009; Reich et al. 2010). One of the distinct advantages of the DLM is that Dzombak and Morel (1990) compiled numerous data sets from previous research and developed a self-consistent data set for hydrous ferric oxide (amorphous iron oxide). This dataset has been used extensively in the lit- erature and in commercially available computer codes such as MINEQL+ and MINTEQA2. Similar attempts to develop databases for other sorbents have emerged in the literature as well (Tonkin et al. 2003). One of the most challenging aspects of modeling natural systems has been to distribute the sorption to different mineral phases. Landry et al. (2009) and Reich et al., (2010)

40 successfully applied the DLM to mineral assemblages contain- ing amorphous iron oxide, quartz and kaolinite. Limitations in the modeling to the quartz component were attributed to failure of the DLM to capture the ionic strength effects. SCMs have also been applied to describe sorption to bac- teria, diatoms, and algae surfaces (Daughney and Fein 1998; Fein et al. 1999; Fowle and Fein 1999; Fein et al. 2001; Yee and Fein 2002; Yee and Fein 2003; Borrok and Fein 2004; Borrok et al. 2004; Borrok et al. 2005; Kaulbach et al. 2005; Gélabert et al. 2006; Alessi and Fein 2010; Deo et al. 2010). Much of this research assumes that the key functional groups on bac- terial surfaces are carboxylic, phosphate, and hydroxyl/amine groups and that surface metal adsorption at acidic and circum- neutral pH is dominated by interactions with the carboxylic groups and utilizes the non-electrostatic surface complexation model. More recently, this assumption has been questioned, and evidence suggests that both carboxylic groups and phos- phate groups may be important (Ngwenya et al. 2009) as well as sulfhydril (Guine et al. 2006; Mishra et al. 2010). In the work of Ngwenya et al. (2009), the constant capacitance model was used to describe lanthanide adsorption to Pantoea agglomerans whereas the non-electrostatic surface complexation model was used in the work of Mishra et al. (2010). 3.5.4 Summary Adsorption to metal ions is arguably the most important process affecting the fate and transport of metal ions. The high affinity of metal oxides for both inorganic and organic surfaces suggests that the major challenge for predicting adsorption is a lack of knowledge regarding the quantity of metal oxide phases in a system. Thus, one of the major challenges for describ- ing adsorption in natural systems is identifying and character- izing the particulate matter within the system. As shown in this review, metal ions adsorb to functional groups on clays, oxide minerals, carbonate solids, bacterial surfaces, algae, fungi, and carbon. In addition, humic and fulvic acids also sorb to these surfaces and impart different surface properties to the particles. These processes must also be considered to understand metal ion sorption. Significant research has helped to elucidate the dominant functional groups and mechanisms of adsorption to surfaces present in natural waters. While the majority of this research has been conducted in well-controlled laboratory studies, research has also shown that the potential for scaling these studies to field scenarios is promising. The development of SCMs has been a key tool that has significantly advanced both understanding of metal ion adsorption and provided reasonable predictions of sorption in complex systems. The refinement and application of these models has proceeded on two different fronts. Many research- ers have focused on developing sophisticated models that accu- rately describe the structure or type of surface complexes, the location of the complexes relative to the surface, and the charge development at the surface from first principles. Others have focused on the application of more simplistic versions of SCMs that capture the trends in metal ion sorption behavior and the extent of sorption as a function of system conditions in com- plex laboratory systems and field systems. The results of these studies have provided the field with the key tools required for predicting metal ion sorption processes; however, in many cases the model parameters may require site-specific calibration. 3.6 Bioavailability and the Biotic Ligand Model Consideration of aqueous speciation and bioavailability is required for predicting metal uptake by microorganisms and understanding metal ion toxicity in aqueous systems (Koster et al. 2004; Van Leeuwen et al. 2005). The fraction of total metal that can be taken up by an organism is termed the bioavailable fraction (Koster et al. 2005). According to Koster et al. (2005), there are three main components that determine the bioavail- ability of a particular chemical species to an organism. These components are the chemical reactivity of the species, the flux or concentration of these species to the microbial surface, and the internalization of the species by the organisms. Figure 3-36 presents a conceptual model of metal-organism interactions (Campbell et al. 2002) that includes these three steps. Even though it is obvious that bioavailability is highly influenced by speciation, it is not clear which metal spe- cies are responsible for the final interaction with the organism. Current models for metal ion uptake by organisms assume that metal ion uptake is proportional to the free metal ion con- centration. While this model is appropriate for many systems, exceptions do occur when metal complexes are transported by an ML ML ML MZ+ MZ+ MZ+ X M L Z Cell interior Plasma membrane Cell wall Diffusion layer Bulk solution Figure 3-36. Conceptual model of metal–organism interactions (adapted from Campbell et al. 2002). Mz, free metal ion; ML, metal complex in solution.

41 independent pathway that does not involve the pathway shown in Figure 3-36 in which dissociation of the complex to the free metal ion precedes uptake by a coordination site on the cell surface. This can include the transmembrane transport of lipophilic complexes [such as Hg(CH3)2] and nanoparticles. Metals such as FeOH3 may be taken up by endocytosis, and metals with hydrated radii similar to common nutrients such as Pb, Cd, Mn, and Co can be taken up by active transport with ATP pumps and calcium channels (Markich and Jeffree 1994; Markich et al. 2001). Over the past decade, the biotic ligand model (BLM) has emerged as the most accepted tool for linking metal ion specia- tion to bioavailability of metal ions (Di Toro et al. 2001). The modeling approach outlined in the BLM has been applied in the scientific community to study bioavailability and metal toxicity (e.g., Villavicencio et al. 2005; Schwartz and Vigneault 2007; Slaveykova 2007), in regulatory communities for establishing water quality criteria (Natale and Leis 2008), and for perform- ing aquatic risk assessments for metals (Janssen et al. 2003). This model is currently used to develop freshwater criteria for copper (U.S. EPA). The goal of the BLM is to successfully predict complexation of the metal ion with the biotic ligand, which is defined as “a specific receptor within an organism where metal complexation leads to acute toxicity” (Santore et al. 2001). The BLM incorpo- rates an equilibrium chemical speciation model that includes competition of the free metal ion with major cations in the sys- tem (e.g., Ca2+, Na+, Mg2+, H+) and complexation by organic and inorganic ligands (e.g., humic substances, chloride, carbonates, and sulfide). The BLM quantitatively relates short-term binding to the biotic ligand to acute toxicity, with lethal accumulation (LA50) as a predictor for 96 h LC50 values for fish or 48 h LC50 values for daphnids (Niyogi and Wood 2004). While the numerous applications of the BLM that appear in the literature are well beyond the scope of this report, a number of key review papers have emerged that summarize the major developments (Di Toro et al. 2001; Paquin et al. 2002; Santore et al. 2002; Batley et al. 2004; Niyogi and Wood 2004; Slaveykova and Wilkinson 2005; Worms et al. 2006). Most importantly for this work, the BLM can consider the effect of solution condi- tions, including pH, major cations such as Ca2+, and competi- tion with solution phase ligands on heavy metal toxicity. The basis of the model is that mortality occurs when the metal- biotic ligand complex reaches a critical concentration. The key assumptions inherent in the development of the BLM are that: CHAPTER 4 Metal toxicity is associated with the free metal ion CHAPTER 5 Complexation with inorganic ligands and NOM decreases metal ion toxicity by decreas- ing free metal ion concentrations CHAPTER 6 Water quality parameters affect metal ion speciation and organism viability CHAPTER 7 Mg and Ca ions that make up hardness reduce toxicity through competition for binding sites on the organism and by stabilizing paracellular junctions in fish gills CHAPTER 8 Different metal ions bind with varying affin- ity to organisms While the original development of the BLM was for fish tis- sue, recent research has revised and applied the model to other organisms such as daphnia (Di Toro et al. 2001; Niyogi and Wood 2004). As a result, the biotic ligand was assumed to be associated with the sodium or calcium channel proteins in fish gill surfaces that regulate the ionic composition of blood. Cur- rent applications to other organisms are modeled in the same way (see Figure 3-37) in that the biotic ligand is assumed to bind to metal cations, compete with other cations, and toxicity occurs when the metal-biotic ligand concentration reaches a critical value. Therefore, the amount of metal that binds is determined by competition for metal ions between the biotic ligand and the other aqueous ligands, particularly NOM, and the competition for the biotic ligand between the toxic metal ion and the other metal cations in solution, for example, calcium. Thus, the model incorporates metal complexation models such as the WHAM Model VI that was described previously. The BLM represents a significant improvement over models that only relate toxicity to the concentration of the free metal cation. A number of recent criticisms of the BLM have emerged over the past several years. Bell et al., (2002) expressed concerns regarding the assumption of equilibrium between all metal Figure 3-37. Representation of the BLM (Di Toro et al. 2001).

42 species in the bulk solution and the biotic ligand. Campbell et al. (2002) and de Schamphelaere et al. (2005) considered assumptions related to internalization which are assumed to be slow relative to the other steps involved in metal uptake, that internalization occurs via cation transport, and that internal- ization must occur for toxicity to appear. Others have focused on the assumption that competing metals can be modeled using an additive approach in which the sites for each metal are independent or the metal ions compete for the sites. However, recent data for Pb and Cu uptake by the green alga Chlamydo- monas reinhardtii demonstrated that Cu plays a synergistic role for the uptake of Pb by aquatic organisms including bacteria (Chen et al. 2010). Thus, further improvements of the model will likely include the potential for different transport systems to dominate under different environmental scenarios including competition for sites, kinetics of internalization as well as other assumptions related to the bioavailability of different chemical species. Recent developments in the BLM include the development of a terrestrial BLM that allows prediction of sediment metal toxicity (Di Toro et al. 2005). The model links measurable sediment and pore water parameters such as simultaneously extracted metals/acid-volatile sulfide (SEM/AVS), pore water- sediment partitioning, and the BLM to produce a median lethal concentration that is normalized to the sediment organic carbon concentration. 3.6.1 Summary The bioavailability of metal ions is dependent on the speciation in solution, the affinity of metal species for surfaces on organisms, the modes of transport through biological membranes, and metal ion toxicity. The BLM represents the state-of-the-art with respect to predicting metal ion toxicity in natural waters. It incorporates chemi- cal speciation models, sorption to the “biotic ligand” of an organism, and toxicity data to evaluate the impact of metal ions on aquatic life. The BLM has significant potential as a regulatory tool. 3.7 Conclusions Chemical speciation in natural waters has significant impli- cations for controlling the mobility, toxicity, and bioavailability of metals in water. Metal ion properties and solution chemis- try dictate the speciation and determine the extent of sorption to inorganic, organic, and biological particulate matter. The key processes that must be considered for predicting metal ion speciation are acid/base chemistry, complexation with simple inorganic ligands, complexation with anthropogenic, autocthonous and NOM, oxidation/reduction reactions, pre- cipitation and sorption. This literature review has presented a brief description of each of these processes and identified their impact on speciation of commonly identified metal ions. The review has identified state-of-the-art approaches used to predict metal ion speciation for each of these processes, and has highlighted interactions among these processes that affect metal ion speciation. The complexity of natural water chemistry was emphasized throughout the review, especially with respect to the heterogeneity of NOM, and approaches for incorporating NOM into predictive speciation models. Finally, the impact of metal ion speciation on bioavailability was addressed, and a description of the BLM was presented. The BLM represents an approach to incorporate all processes affect- ing metal ion speciation into a toxicity model for predicting the impact of metal ions on target organisms (the biotic ligand). The literature review demonstrates that the BLM has significant potential as a tool for incorporating metal ion speciation into bioavailability and toxicity assessment. Average concentrations of trace elements in freshwaters, stormwater, and highway runoff range over orders of mag- nitude, and their mobilities in these systems vary depending on the metal ion, the chemical speciation, and the extent of partitioning. However, in most natural systems only the most mobile elements (e.g., Na, B, Se, As) will be transported in the dissolved phase. Thus, partitioning to particulate and colloi- dal matter is the key to understanding fate and transport in water. The composition of colloidal matter in natural water includes highly reactive humic and fulvic acids which impact metal ion speciation directly through complexation and indi- rectly through competition for adsorption sites. The research suggests that much of the reactivity of humic and fulvic acids is due to the presence of carboxylic (-COOH) and phenolic (C6H5OH) groups. Thus, models that describe metal ion spe- ciation in waters must incorporate parameters that describe reactions with these functional groups as well as those that occur with inorganic ions and particulate phases. Speciation of the dissolved metal ion fraction is a function of the pH, ionic strength, and presence of other ions, organic molecules, and solid phases. Complexation reactions prevent metal ions from being precipitated, complexing agents act as carriers for trace elements in water, and metal ion bioavailabil- ity is often reduced by complexation. Simple inorganic ligands such as carbonate can control speciation in natural waters only in the absence of significant concentrations of stronger organic ligands. Naturally occurring and anthropogenic chelating agents that have binding constants that are orders of magni- tude higher than the simple inorganic and organic acid func- tional groups can also dictate speciation if present at sufficient concentration. Because metal-ligand complexes can have sig- nificantly different charge than the metal ion itself, the mobil- ity of these species can be increased dramatically. For example, Cu-EDTA species are typically negatively charged and highly mobile. Chemical speciation models provide accurate tools

43 for predicting chemical speciation in water. The thermo- dynamic constants for complexation of metal ions with inorganic ligands, organic acids, and well-characterized anthro- pogenic ligands have been incorporated into databases that are tied to these programs. However, the approaches used to include the functional groups associated with humic and ful- vic acids are variable, and the thermodynamic constants have significantly higher uncertainty. While general trends in complexation with NOM have been observed, complexation with NOM is more difficult to define due to the complex and ill-defined nature of these organic complexes and the variability in the number and types of functional groups. Challenges associated with describing complexation with NOM still remain. Recent studies suggest that the reactive components of NOM can be characterized using a series of simple organic acids that range from weakly binding carboxylate and phenolate to stronger binding salicy- late, citrate, and phthalate molecules. In general, complexation of these functional groups with metal cations increases with pH, decreases with metal ion concentration, and increases with fulvic and humic acid concentration. Several models have emerged over the past several decades, and the two most promising models that capture these general trends including the NICA-Donnan model and Model VI. These models have demonstrated ability to predict metal ion binding over a range of conditions and represent the current state-of-the-art. Several metals including Cr, Se, Co, Pb, As, Ni, and Cu are redox active in natural systems and can exhibit different lev- els of toxicity, mobility, and bioavailability depending on the oxidation state of the metal ion. The particular oxidation state of a metal is typically controlled by dominant biogeochemical processes occurring within the system. Several metals such as Cr and As exhibit multiple oxidation states in natural waters and the relative rates of oxidation/reduction are often controlled by interactions with NOM and surfaces. In addition, metal fate and transport is affected by the dominant redox pro- cesses that affect speciation of major anions such as nitrate and sulfate. In reducing environments, nitrate reduction to ammo- nia increases metal ion complexation and sulfate reduction to sulfide can lead to precipitation of many of the trace metal ions. Incorporation of redox chemistry has also been included in most geochemical codes; however, in many cases these pro- cesses are rate limited, and thermodynamic approaches are not appropriate. Precipitation processes also have a large impact on the specia- tion of metal ions in solution. In oxic waters, the major precipi- tates include hydroxide and carbonate phases. While published solubility constants are available for most of the potential solid phases, significant variability exists in the literature due to the experimental methods and conditions used to determine the values. Moreover, precipitation processes are often rate limited, and the most stable phase is typically not the dominant phase found in natural waters. The kinetics of precipitate formation are dependent on the presence of nucleation sites from other similar phases or from homogenous solid formation. The evi- dence is also mounting that removal of metal ions in natural and engineered systems is due to the more rapid formation of mixed precipitates or co-precipitates formed from either dis- solution of prevailing solid phases (e.g., clays and oxide miner- als) or precipitation of more insoluble phases such as iron and aluminum hydroxide. Removal of metal ions from solution may also be controlled by adsorption to these other precipi- tating phases as is commonly noted in water treatment appli- cations. Finally, the importance of nanoparticle formation is emerging as a key process controlling the fate of a number of metal ions such as Zn and Hg. This area of research is only in its infancy, but has the potential to impact both the thermo- dynamics and kinetics of metal ion speciation. Adsorption to metal ions is arguably the most important process affecting the fate and transport of metal ions. The high affinity of metal oxides for both inorganic and organic surfaces suggests that the major limitation to adsorption is the availabil- ity of surfaces. Thus, one of the major challenges for describing adsorption in natural systems is identifying and characteriz- ing the particulate matter within the system. As shown in this review, metal ions adsorb to functional groups on clays, oxide minerals, carbonate solids, bacterial surfaces, algae, fungi, and carbon. Adsorption to oxide and carbonate minerals is highly pH dependent, with increasing adsorption at increasing pH for cation metal ion adsorption to oxide minerals. For strongly sorbing metal ions such as Pb, Zn, and Cu, ionic strength has only minimal impacts on the extent of adsorption. Competi- tion among solutes can also have a major impact on adsorption, but in many cases the presence of oppositely charged metal ion species (e.g., divalent cations and oxyanions) can increase adsorption of both species through surface charge reduction or ternary species formation. In addition, humic and fulvic acids also sorb to these surfaces and impart different surface proper- ties to the particles. These processes must also be considered to understand metal ion sorption. Significant research has helped to elucidate the dominant functional groups and mechanisms of adsorption to surfaces present in natural waters. While the majority of this research has been conducted in well-controlled laboratory studies, research has also shown that the potential for scaling these studies to field scenarios is promising. The development of SCMs has been a key tool that has significantly advanced both our understanding of metal ion adsorption and provided reasonable predictions of sorption in complex systems. The refinement and application of these models has proceeded on two different fronts. Many research- ers have focused on developing sophisticated models that accu- rately describe the structure or type of surface complexes, the location of the complexes relative to the surface, and the charge

44 development at the surface from first principles. Others have focused on the application of more simplistic versions of SCMs that capture the trends in metal ion sorption behavior and the extent of sorption as a function of system conditions in complex laboratory systems and field systems. The results of these studies have provided the field with the key tools required for predicting metal ion sorption processes; however, in many cases the model parameters may require site-specific calibration. The bioavailability of metal ions is dependent on the spe- ciation in solution, the affinity of metal species for surfaces on organisms, the modes of transport through biological membranes, and metal ion toxicity. The BLM predicts com- plexation of the metal ion with a specific receptor site within an organism (the biotic ligand) where metal complexation leads to acute toxicity. The BLM represents the state-of-the- art with respect to predicting metal ion toxicity in natural waters. It incorporates chemical speciation models, sorption to the “biotic ligand” of an organism, and toxicity data to evaluate the impact of metal ions on aquatic life. The BLM has significant potential as a regulatory tool. While the complexity of metal ion speciation hampers the a priori assessment of metal ion mobility, toxicity, and bio- availability, the availability of predictive models provides a set of tools that allow assessment of these if the key parameters of the system have been quantified. These parameters include the pH, ionic strength, organic carbon content and compo- sition, the presence of competing and complexing solutes, and the composition and quantification of particulate and colloidal matter within the system. In many cases, the appli- cation of the most sophisticated predictive models is not suit- able due to the lack of data describing the system. However, the literature suggests that even simplistic versions of these models that capture the general trends associated with key parameters such as pH and complexation may be adequate for many applications.