Loading...
HomeMy WebLinkAboutNC0000396_App C Geochemical Model_20160219Revision 0, 1/27/2016 Analysis of Geochemical Phenomena Controlling Mobility of Ions from Coal Ash Basins Brian A. Powell, Ph.D. 112 Cherry Street Pendleton, SC 29670 (864) 760-7685 bpowell@clemson.edu January 27, 2016 Revision 0, 1/27/2016 TABLE OF CONTENTS EXECUTIVE SUMMARY .............................................................................................................. i  1. INTRODUCTION ................................................................................................................ 1  2. GEOCHEMICAL MODEL DEVELOPMENT .................................................................... 2  2.1. Sorption model development ............................................................................................ 2  2.2. Geochemical model parameterization ............................................................................... 6  2.3. Pourbaix diagram modeling ............................................................................................ 15  2.4. The use of Kd values ....................................................................................................... 16  3. GEOCHEMICAL MODELING OF ARSENIC ................................................................. 18  3.1. Pourbaix diagram analysis .............................................................................................. 18  3.2. PHREEQC model analysis .............................................................................................. 19  3.3. Comparison of measured and calculated As speciation .................................................. 26  3.4. Comparison modeled and experimental Kd values for arsenic ........................................ 29  4. GEOCHEMICAL MODELING OF BORON .................................................................... 30  4.1. Pourbaix diagram analysis .............................................................................................. 30  4.2. PHREEQC speciation analysis ....................................................................................... 30  4.3. Comparison between modeled and experimental Kd values for boron ........................... 32  5. GEOCHEMICAL MODELING of CHROMIUM .............................................................. 34  5.1. Pourbaix diagram analysis .............................................................................................. 34  5.2. PHREEQC speciation analysis ....................................................................................... 34  5.3. Comparison between modeled and experimental Kd values for chromium .................... 39  6. GEOCHEMICAL MODELING of MANGANESE ........................................................... 41  6.1. Pourbaix diagram analysis .............................................................................................. 41  6.2. PHREEQC speciation analysis ....................................................................................... 43  6.3. Comparison between modeled and experimental Kd values for manganese ................... 45  7. SUMMARY ........................................................................................................................ 46  8. REFERENCES ................................................................................................................... 48  Page i EXECUTIVE SUMMARY The goal of this geochemical modeling effort is to describe the geochemical behavior and subsurface mobility of several constituents of interest in the subsurface by considering sorption of the constituent to the aquifer solids, oxidation/reduction reactions, and precipitation/coprecipitation in mineral phases using the United States Geological Survey program PHREEQC. This report describes the geochemical behavior relevant to seven coal ash basin storage sites which will be referred to in this report by the abbreviated plant names: HF Lee, Weatherspoon, Mayo, Cape Fear, Sutton, Asheville, and Roxboro. A major effort was undertaken to describe the chemical speciation expected under the variable conditions (particularly with respect to changes in pH and EH) and to relate the expected speciation to observed behavior of each constituent. The model also considers the influence of background major ion concentrations on the sorption of constituents of interest to iron and aluminum hydroxide solid phases. To provide a self-consistent set of thermodynamic constants for sorption reactions, all sorption was modeled assuming hydrous ferric oxide (HFO) and hydrous aluminum oxide (HAO) were the dominant sorbing surfaces based on the databases developed by Dzomback and Morel [1] and Karamalidis and Dzomback [2]. The model input requires initial concentrations of all ions of interest, pH, redox potential (EH), as well as a concentration of sorption sites. The field data from seven sites was compared and it was demonstrated that each of these parameters were relatively consistent between the seven sites. Therefore, an average range of geochemical conditions were used to predict the minimum and maximum distribution coefficients (Kd) for each constituent of interest. The initial concentrations of ions were determined by examining the minimum, average, and maximum concentrations observed by compiling the data from all seven sites under consideration. A range of pH and EH values was selected to capture the range of conditions observed at all seven sites. Selection of an appropriate pH and EH range is vital because these two variables have the greatest impact on constituent Kd values. Thus the pH and EH values selected for the model represented a wide range capturing minimum and maximum values. The concentration of iron (Fe) and aluminum (Al) sorption sites was estimated based on the average extractable iron and aluminum content of the solid phases retrieved from all seven sites. Parameterization of the model in terms of the sorption site concentrations has a direct and linear relationship to the predicted Kd values. The average extractable iron and aluminum concentrations from the solid samples was converted to a sorption site concentration by assuming site densities of 0.2 moles of Fe sites per mole of solid phase Fe and 0.4 moles of Al sites per mole of solid phase Al. Essentially assuming an increase in the iron or aluminum site density will cause a corresponding increase in the sorption site concentration. The major emphasis in describing the predicted Kd values should be on comparing the trends with changing system parameters (i.e. EH, pH, ion concentrations) rather than comparing specific Kd values. The partitioning and solubility of constituents is highly dependent on the pH of the ground water. This is because the majority of constituents of interest exist as anionic or cationic species. Sorption of charged species to mineral surfaces changes with pH because the surface charge of all mineral surfaces transitions from a positively charged surface at low pH to a negatively charged surface at high pH. Therefore, sorption of anionic species will be stronger at low pH where the anions are attracted to the positively charged surfaces (vice versa regarding the cationic species). Similarly, the solubility of a mineral phase will also be pH dependent because lower pH values tend to favor the formation of more soluble cationic species of most alkali elements, alkali earth elements, and transition metals. Conversely, low pH values will facilitate protonation of most oxoanions (such as the conjugate bases AsO43-, SeO32-) which can form neutrally charged H3AsO4 and H2SeO3 species at low pH. At higher pH values, these Page ii oxoanions deprotonate and persist as anionic species which are generally very soluble and will only weakly sorb to mineral surfaces. Therefore, generally low pH conditions will favor higher aqueous concentrations of cationic constituents (e.g., Ba2+, Cr3+, Co2+, Fe2+/Fe3+) whereas higher aqueous concentrations of anionic species (e.g., AsO43-, SeO32-, H2VO42-, H2BO3-) will be expected in higher pH ground waters. Since the partitioning of these constituents is highly dependent on the pH and the chemical speciation of the constituent, consideration of potential changes in the constituent chemical species due to changes in oxidation state is imperative. For example, Cr(III) generally exists as the cation Cr3+ which is relatively insoluble and sorbs strongly to mineral surfaces. However, upon oxidation to Cr(VI), the oxyanion chromate CrO42- becomes the dominant species which is highly soluble and mobile under neutral to high pH conditions. The geochemical model developed in this work considers changes in oxidation state for all redox active constituents of interest (Se, As, Fe, V, Mn, Cr, Co, S) and changes in chemical speciation for all constituents. Some specific observations are as follows:  Arsenic: The PHREEQC model predicts As(V) as the dominant oxidation state of arsenic under the field measured EH and pH conditions but As(III) is the dominant species measured in ground waters. The reason for this discrepancy is proposed to be due to 1) increased sorption of As(V) relative to As(III) which would remove all As(V) from the ground water and prevent As(V) measurements in samples and/or 2) a kinetic limitation with respect to the As(III)/As(V) oxidation/reduction reactions which prevents the system from reaching chemical equilibrium. However, the observation of As(III) is consistent with the relatively lower Kd values required in the reactive transport modeling efforts compared with the higher Kd values predicted by PHREEQC. Therefore, the reactive transport model represents a conservative estimate. Due to the stronger sorption of As(V), the tendency of the element to move in the subsurface, will decrease as As(III) becomes oxidized to As(V) and sorbs to mineral surfaces. Additionally, the minerals scorodite (FeAsO4.2H2O) and mansfieldite (AlAsO4.2H2O) are near saturation under some pH and EH conditions examined in this model and measured in the field. Thus these minerals may theoretically form but generally are unlikely mineral phases to form in the shallow subsurface.  Boron: Boron exists only in the B(III) oxidation state and generally persists as the neutrally charged chemical species boric acid (H3BO3), which is a weak acid and exhibits minimal sorption to mineral surfaces. As the system pH increases, H3BO3 will deprotonate (i.e. release a H+ ion) to form H2BO3- which also sorbs weakly. Boric acid and H2BO3- are the only two aqueous species of boron predicted to occur in this model. Thus, the PHREEQC predicted Kd values for boron are low (1.1 x 10-5 to 0.34 L/kg). These values are slightly lower but generally consistent with the values chosen for reactive transport modeling [3-9] and those measured in batch laboratory experiments [10-16]. Precipitation of any boron containing mineral phases is not expected to occur. Therefore, physical attenuation and sorption are the two primary processes which will control the movement of boron in the subsurface.  Chromium: The ground water measurements from Weatherspoon indicate Cr(III) is the dominant oxidation state which is in agreement with the PHREEQC model. The sorption of Cr(III) is significantly stronger than Cr(VI) because Cr(III) persists as a highly charged cation (Cr3+) which readily sorbs to mineral surfaces as the pH increases from acidic to basic conditions. This behavior is in stark contrast to that of Cr(VI) which persists as a weakly sorbing anion (CrO4-) and decreases sorption from acidic to basic conditions. This high charge density of Cr3+ also Page iii causes a propensity to form aqueous complexes with anions such as SO4-2 and Cl- which can influence sorption behavior. For example, formation of CrSO4+ appears to be responsible for a decreased Kd relative to baseline conditions in the PHREEQC model presented in this work. The measured aqueous concentrations in groundwater from the seven sites range from below detection to approximately 100 g/L. This concentration range is similar to what was modeled in PHREEQC and is indicates that formation of mineral phases containing Cr may occur under high pH conditions with relatively high Cr concentrations.  Manganese: Manganese is predominantly present as Mn2+ in the PHREEQC model output which is in agreement with the measurements of Mn(II) in groundwater from the Weatherspoon, Lee, and Sutton sites. Sorption of Mn(II) is generally weak and yields Kd values ranging from 5 x 10-7 to 5 L/kg calculated from the PHREEQC model. The mean Kd value of 7 x 10-3 L/kg and the Kd value of 0.15 L/kg for the average groundwater conditions from the PHREEQC model are in reasonable agreement with the range of 0 to 0.10 L/kg used in reactive transport models [3, 4, 8] and the value of 9 x 10-3 L/kg determined from batch sorption experiments with solids from the Cape Fear site. Analysis of the saturation index of mineral phases containing Mn from the PHREEQC model indicates that several common Mn bearing soil minerals (manganite, hausmannite, and birnessite) are near saturation under high pH and high EH conditions. Therefore, precipitation of Mn mineral phases could occur given sufficiently high Mn concentrations and high pH/EH conditions. Based on the above analysis, it is expected that physical attenuation, sorption, and chemical precipitation could all play a role on controlling the movement of Mn in the subsurface. Page 1 1. INTRODUCTION A geochemical modeling effort was undertaken to describe the chemical speciation expected under the variable conditions at seven Duke Energy Progress power plant sites: Sutton, Weatherspoon, H.F. Lee, Mayo, Cape Fear, Asheville, and Roxboro. The primary emphasis of the geochemical modeling effort was to understand the influences of pH and redox potential (EH) on the aqueous speciation, sorption, and solubility of several constituents of interest using the United States Geologic Survey (USGS) geochemical modeling program PHREEQC. In the previous models, a wide range of pH and EH conditions at each site was modeled using site relevant data [17-23]. Hydrous ferric oxide (HFO) and gibbsite (HAO) minerals were used as the basis for sorption and capacity determination because of the available thermochemical databases for surface complexation modeling of many constituents of interest [1, 2]. This model assumed sorption occurred only to iron oxide surfaces based on sorption reactions to HFO described by Dzomback and Morel [1]. In this revised model, sorption to HAO was also modeled based on the reactions compiled by Karamalidis and Dzombak [2]. The concentrations of HFO and HAO were constrained based on extractable Al and Fe concentrations from solid phases recovered from each site measured by collaborators at the University of North Carolina – Charlotte [10-16]. The approach taken in this “global” modeling effort was to understand how changes in pH, redox potential, and dissolved ion concentrations influence the sorption, aqueous speciation, and solubility of several constituents of interest. The pH, EH, and ion concentrations from all seven sites are compared below to demonstrate that they are all relatively similar. Therefore, a fixed range of values was used to perform the geochemical modeling discussed in this report. The logic of this model is that it is essentially impossible to predict a Kd from first principles to use in a reactive transport model considering the multitude of chemical, physical, and potentially biological processes occurring at the coal ash basin sites. Therefore, the primary emphasis was to quantify how changes in the system conditions will alter the speciation and mobility of each constituent (particularly changes in pH and EH). This will allow us to determine if changes occurring during remediation could mobilize any particular constituent. Page 2 2. GEOCHEMICAL MODEL DEVELOPMENT 2.1. Sorption model development To examine the sorption behavior of multiple ions of interest in these systems, a combined aqueous speciation and surface complexation model was developed using the USGS geochemical modeling program PHREEQC. Equilibrium constants for aqueous speciation reactions were taken from the USGS WATEQ4F database. This database contained the reactions for most elements of interest except for Co, Sb, V, and Cr. Constants for aqueous reactions and mineral formation for these elements were taken from the MINTEQ v4 database which is also issued with PHREEQC. The constants were all checked to provide a self-consistent incorporation into the revised database. The source of the MINTEQ v4 database is primarily the well-known NIST 46 database [24]. Sorption reactions were modeled using a double layer surface complexation model. To ensure self-consistency in the sorption model, a single database of constants was used as opposed to searching out individual constants from the literature. The diffuse double layer model describing ion sorption to HFO and HAO Dzomback and Morel [1] and Karamalidis and Dzombak [2], respectively, was selected for this effort. Many surface complexation reactions for ions of interest on HFO are included in the standard release of the PHREEQC database. Constants for Co, V, Cr, and Sb were added to the modified database as well as all constants involving ion sorption to HAO. Using surface complexation models, the sorption of an element is written as a standard chemical reaction such as those shown in Table 2.1. In these equations, ≡SOH represents a site on the HFO or HAO mineral surface where sorption can occur. Speciation models utilize this reaction convention to describe a “concentration” of surface sites to be used in a thermochemical approach to sorption modeling [1, 25-27]. The primary difficulty in this approach is quantifying the concentration of reactive surface sites. Many approaches have been used, the most common being potentiometric titrations of the solid phase to quantify surface site concentrations using proton sorption/desorption behavior and surface area analysis. These studies are typically done on pure, synthetic mineral phases and still exhibit large variations in the surface site density determined from the data. Therefore, determination of surface site densities for complex mineral assemblages cannot be accurately performed using currently available techniques. To constrain the number of sorption sites to be used in this model, a concentration of surface sites (SOH) must be calculated in units of mol/L for application in the aforementioned chemical equations. Such a concentration is conceptually difficult because SOH represents a point on a solid particle where another ion may sorb, not an aqueous species as indicated by the units of mol/L. So to make this transition, a density of sorption sites on the mineral surface must be assumed (e.g. “x” moles of sorption sites per mole of total iron or aluminum in the solid phase). Additionally, in order to calculate a Kd value to compare with batch laboratory data, a solid phase concentration in gsolid/L must also be assumed. Page 3 Table 2.1: Example reactions used in surface complexation modeling (where SOH represents a sorption site). Reaction Type Reaction Expression Stability constant Surface protonation (i.e. develops positive surface charge at low pH) SOH + H+  SOH2+ Surface deprotonation (i.e. develops negative surface charge at high pH) SOH  SO- + H+ Cation sorption SOH + Mn+  SOMn-1 + H+ Anion Sorption SOH + H+ + A-  SOH2+A- or SOH + A-  SA + OH- K୅ ൌ ሾSOHଶା Aି ሿ ሾSOHሿሾH ା ሿሾAି ሿ exp ൬െ ܨΨ ܴܶ ൰ However, this assumed value essentially is canceled out when back calculating a Kd value from the geochemical modeling data. Therefore, the concentration of sorption sites based on extractable Fe concentrations is based on the following equation: ሾ≡ ܨܱ݁ܪ ሿ ൌ ሾܵ݋݈݅݀ሿ ∗ሾܨ݁ሿ௘௫௧௥ ∗1݃ி௘ 1000݉݃ி௘ ∗݉݋݈ி௘ 55.845݃ி௘ ∗0.2 ݉݋݈ ≡ ܨܱ݁ܪ ݉݋݈ ܨ݁௘௫௧௥௔௖௧௔௕௟௘ where [Solid] is the solid phase suspension concentration in g/L assumed for the model, [Fe]extr is the extractable Fe concentration in mgFe/gsolid, and [≡FeOH] is the concentration of iron surface sites in the model input. However, the model output is in mol/L of sorbed ions and mol/L of aqueous ions. Therefore to convert to a Kd value, we must convert the mol/L of sorbed ions to mol/kgsolid. This is done by using the solid phase concentration assumed in the above reaction to keep the model self-consistent and essentially cancel the assumed solid phase concentration as noted above. This is the approach used in this model with an assumed solid phase concentration of 50 g/L. An alternative approach when the site density is either known or assumed (i.e. sites per nm2 of mineral surface area), a molar concentration of surface sites can be determined using the equation: ሾ≡ ܨܱ݁ܪ ሿ ൌ ሾܵܵሿ ∗ܵܣ∗10ଵ଼ ݊݉ଶ ݉݉ଶ ∗ܵܦ 6.022 ݔ 10ଶଷ ሺܽݐ݋݉ݏ ݉݋݈ ሻ + 2 ++ [SOH ]FψK= exp[SOH]{H } RT  -+ - [SO ]{H } FψK= exp -[SOH] RT  + n-1 + M [SOM ]{H } FψK= exp( 1)[SOH]{M } RT n n  Page 4 where [≡FeOH] is the concentration of iron sorption sites in mol/L, [SS] is the suspension of solids in g/L, SA is the surface area of the solid in m2/g, and SD is the site density of the solid (sites/nm2). The model proposed by Dzomback and Morel [1] assumes that all surfaces have a combination of strong sorption sites and weak sorption sites. As discussed above, quantifying the reactive surface site density for complex mineral assemblages such as those used in this work, is difficult if not impossible. Therefore, attempting to delineate between mineral surfaces, let alone strong and weak sites on such surfaces, would add unnecessary uncertainty and fitting parameters to the models. Therefore, sorption to only one site on both HFO and HAO is considered. There are two primary approaches to modeling complex mineral assemblages such as those considered in this work. The component additivity approach considers sorption reactions to all mineral phases present in a sample [25]. Such an approach requires separate reactions for each analyte sorbing to each mineral phase present in a sample. These can be very complicated but robust models provided a means for determining the surface site density of each mineral phase is available. A simpler alternative is the generalized composite approach wherein data are modeled assuming a generic surface site (i.e.,≡SOH) which represents an average reactivity of all minerals in the solid assemblage [25]. This modeling approach still combines the flexibility of an aqueous speciation model with a sorption model under a thermochemical framework. This work assumes that sorption occurs only to iron oxide minerals. Other mineral surfaces can be considered and modeled. However, in the absence of data with sufficient resolution to determine the presence of these mineral phases and accurate methods to determine the surface site density for the minerals being considered, fitting additional surface reactions becomes a curve fitting exercise with a high probability of a non-unique solution. By modeling ion sorption to HFO and HAO based on extractable metal content but not considering other phases, the model is essentially a combined generalized composite and component additivity model. An average sorption site concentration of HFO and HAO for the model was determined by comparing the extractable Al and Fe concentrations in solids from all seven sites. The average concentrations of extractable Al and Fe from solids obtained from all seven sites are shown in Figure 2.1. These data indicate that the average concentrations are relatively similar though there is a significant amount of variation at each site. To see the data in finer resolution, the minimum, mean, average, and maximum values are shown in Figure 2.2. Due to the relatively similar values of extractable Fe and Al at each site, a “global” sorption model was selected using the average extractable Fe and Al from the data from all seven sites. This average was used to determine the input concentration of sorption sites for the PHREEQC model. Table 2.2: Average extractable Fe and Al concentrations from all sites and calculated molar site concentrations for PHREEQC model. Values are based on extractable Fe and Al measurements reported by Langley and Shubhasini [10-16]. mgFe-Al/ kgsolid molFe-Al/ gsolid molsites/ gsolid Assuming 50 gsolid/L, site concentration in mol/L Extractable Fe 1002 1.79E-05 3.59E-06 1.79E-04 Extractable Al 762 2.83E-05 1.16E-05 5.79E-04 Page 5 In previous geochemical models, the site densities of HAO and HFO sorbents were used to determine the capacity of the solid phases at each of the seven sites. Using the same method described above, site densities available for sorption were calculated based on the average extractable Fe and Al concentrations from each site [28-34]. Then modeling groundwater concentrations of each constituent of interest at the NC2L Standard Level and conservatively assuming 100% sorption, the capacity of the solid phases to sorb the constituents of interest was determined. In all cases, less than 1% of the total sorption capacity of the solid phases was occupied by the constituents of interest [28-34] . Therefore, solid phases from each site were shown to have the capacity to sorb all of the available constituents of interest considering the above assumptions. However, it is important to note that these calculations assume 100% sorption which will not be the case for constituents with low Kd values such as boron and manganese. Therefore, although this calculation shows that it is unlikely the capacity of the aquifer solids would be exceeded, the actual aqueous phases of each constituent would be based on the Kd value for that constituent under the geochemical conditions of the pore water. A similar result occurs when using the average sorption site concentrations from the average extractable Fe and Al concentrations listed in Table 2.2 and used in this report. Therefore, in the PHREEQC model output discussed in this report, saturation of available surface sorption sites is not expected to occur. Figure 2.1: Average extractable Al and Fe concentrations in solids from all seven sites. 0 500 1000 1500 2000 2500 3000 3500 4000 Ex t r a c t a b l e A l o r F e ( m g / k g so l i d ) Average Extractable Fe Average Extractable Al Page 6 Figure 2.2: Minimum, mean, and maximum extractable Fe (top) and Al (bottom) concentrations in solids from all seven sites. Based on data from [10-16] 2.2. Geochemical model parameterization The main geochemical parameters influencing sorption are pH, EH, and the availability of sorption sites. The pH and EH of numerous ground water samples has been measured at each of the seven sites along with other relevant geochemical parameters including the dissolved ion concentrations and the oxidation state speciation of redox active ions such as As, Cr, and Se [17-23]. Similar to the method described above to examine the similarities of extractable Fe and Al concentrations from solid phases, the measured pH and EH values from the seven sites are compared in Figure 2.3. 0 1000 2000 3000 4000 5000 6000 7000 MIN GEOMEAN AVG MAX Ex t r a c t a b l e F e ( m g Fe /k g so l i d ) Asheville Sutton Lee Mayo Roxboro Cape Fear Weatherspoon 0 1000 2000 3000 4000 5000 6000 7000 8000 MIN GEOMEAN AVG MAX Ex t r a c t a b l e A l ( m g Al /k g so l i d ) Asheville Sutton Mayo Roxboro Cape Fear Weatherspoon Page 7 Figure 2.3: Measured pH and EH values at all seven sites. The dashed lines represent the conditions where water is stable (i.e. below the bottom line water will reduce to H2(g) and above the top dashed line water will oxidize to O2(g)). The inset box represents an approximate range of pH and EH values that will capture the majority of conditions of the site. The open symbols represent the values selected for PHREEQC modeling. Based on data from [17-23]. The values from each site all fall within a similar range and there is not a particular site with drastically different values than the others. Therefore, a reasonable range of pH and EH values was selected for all sites to parameterize the speciation model which will cover the expected range of chemical speciation expected at the sites. It is a reasonable assumption that if the speciation models are run using the approximate EH and pH conditions within the inset box of Figure 2.3, the geochemical behavior of each constituent can be determined. The variability of the pH and EH conditions at each site will essentially be “noise” considering the wide range of Kd values predicted as a function of pH which are discussed below. Therefore, one “global” model which shows the influence of Kd as a function of pH and EH within the selected range is appropriate for all sites. The range of selected pH and EH values are depicted in Table 2.3 with the black open circles. This range was chosen by selecting values to represent a global average, a high pH/low EH extreme, a high pH/high EH extreme, a low pH/low EH extreme, a low pH, high EH extreme, and values representing the pH at 25% and 75% cumulative fractions from the histograms in Figure 2.4 and 2.5. Such a model could also be used to delineate between geologic units and choose the Kd values which are best represented by the EH and pH values of that unit. However, the reactive transport modeling -350 -150 50 250 450 650 24681012 E H (m V ) pH Weatherspoon Sutton Lee Cape Fear Asheville Mayo Roxboro Water Oxidation Water Reduction Selected Values Page 8 and similarity in the pH and EH conditions between the different geologic units at each site do not necessarily justify this effort. Thus, the approach taken in this report is to compare the Kd values required by the reactive transport model, Kd values from the PHREEQC geochemical model, and Kd values measured in laboratory experiments to ensure that the trends, which are a manifestation of the underlying geochemical behavior, are similar. Figure 2.4: Histograms of EH values measured at all seven sites. The bottom figure shows the mean EH values at +/- 1 standard deviation in the blue and black lines, respectively. Values at approximately 25% and 75% for each for each pH and EH condition were included in geochemical modeling at each site (Table 2.3). Based on data from [17-23]. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 10 20 30 40 50 60 -400 -300 -200 -100 0 100 200 300 400 500 600 700 800 Cu m u l a t i v e F r a c t i o n # o f O c c u r r e n c e s EH (mV) Histogram Cumulative Fraction 0 10 20 30 40 50 60 70 -400 -300 -200 -100 0 100 200 300 400 500 600 700 800 # o f O c c u r r e n c e s EH (mV) Histogram EH + 1 SD EH mean value EH - 1 SD Page 9 Figure 2.5: Histograms of pH values measured at all seven sites. The bottom figure shows the mean pH values at +/- 1 standard deviation in the blue and black lines, respectively. Values at approximately 25% and 75% for each for each pH and EH condition were included in geochemical modeling at each site (Table 2.3). Based on data from [17-23]. Table 2.3: pH and EH Values for Global Model Input. Note that EH was entered in PHREEQC using pe based on the equation EH = 59 mV * pe pH EH (mV) pe Notes 4 482 8.16 low pH, high EH value 5.6 -21 -0.35 low pH, low EH value 6.47 220 3.72 global average pH and EH 6.9 514 8.69 high pH, high EH value 9.1 -104 -1.75 high pH, low EH value 5.1 372 6.29 pH range covering 25-75% of sites from Figure 2.4 7.1 75.5 1.28 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 5 10 15 20 25 30 35 40 45 2 3 4 5 6 7 8 9 10 11 12 Cu m u l a t i v e F r a c t i o n # o f O c c u r r e n c e s pH Histogram Cumulative Fraction 0 5 10 15 20 25 30 35 40 45 50 23456789101112 # o f O c c u r r e n c e s pH Histogram pH + 1 SD pH mean value pH - 1 SD Page 10 In addition to the pH and EH range, a range of ion concentrations must also be selected for the PHREEQC modeling. Similar to the model parameterization discussed above, the measured values from all seven sites were compared to determine if there was significant variability. Hundreds of data points from various geologic units at all seven sites were plotted together in the following series of figures (Figures 2.6-2.8). There are wide variations in the ion concentrations at each site, but the average values from site to site are relatively constant. Therefore, it was assumed that a global set of average values could be used to approximate the geochemical behavior at each site. Using these values, a set of three conditions were used as input values for the model (Tables 2.4 and 2.5). The concentrations of major ions (e.g., Ca2+, Na+, Fe(II/III), Cl-, SO42-) were varied to consider the range of potential values. The concentrations of several trace ions and constituents of interest were not varied so that the model could examine the potential for competition for sorption sites between the varying major ion concentration conditions and a fixed condition for the trace elements. For example, sorption of Fe2+ and SO42- can effectively block sites for cation and ion sorption, respectively. Therefore, by considering the ranges of ferrous iron and sulfate listed below in Table 2.5, the potential for sulfate to outcompete another anion (e,g., arsenate AsO4-3) can be examined. In the model output described below, this impact is demonstrated by comparison of the Kd values measured under “low”, “average” and “maximum” groundwater ion concentrations based on the values in Tables 2.4 and 2.5. Page 11 Figure 2.6: Minimum, average, and maximum Al, Sb, As, Ba, B, Cd, and Cr concentrations in groundwater at all seven sites. Based on data from [17-23]. 0.1 1 10 100 1000 Mi n i m u m C o n c e n t r a t i o n ( u g / L ) Sutton Weatherspoon Lee Cape Fear Roxboro Mayo Asheville 1 10 100 1000 10000 Av e r a g e C o n c e n t r a t i o n ( u g / L ) Sutton Weatherspoon Lee Cape Fear Roxboro Mayo Asheville 1 10 100 1000 10000 100000 Ma x i m u m C o n c e n t r a t i o n ( u g / L ) Sutton Weatherspoon Lee Cape Fear Roxboro Mayo Asheville Page 12 Figure 2.7: Minimum, average, and maximum Cu, Pb, Mn, Mo, Ni, Se, Sr, and V concentrations in groundwater at all seven sites. Based on data from [17-23]. 0.001 0.01 0.1 1 10 100 Mi n i m u m C o n c e n t r a t i o n ( u g / L ) Sutton Weatherspoon Lee Cape Fear Roxboro Mayo Asheville 0.1 1 10 100 1000 10000 Av e r a g e C o n c e n t r a t i o n ( u g / L ) Sutton Weatherspoon Lee Cape Fear Roxboro Mayo Asheville 1 10 100 1000 10000 100000 Ma x i m u m C o n c e n t r a t i o n ( u g / L ) Sutton Weatherspoon Lee Cape Fear Roxboro Mayo Asheville Page 13 Figure 2.8: Minimum, average, and maximum Ca, carbonate, Cl, Fe, Mg, nitrate, K, Na, sulfate, sulfide, and Zn concentrations in groundwater at all seven sites. Based on data from [17-23]. 0.01 0.1 1 10 100 1000 Av e r a g e C o n c e n t r a t i o n ( m g / L ) Sutton Weatherspoon Lee Cape Fear Roxboro Mayo Asheville 0.001 0.01 0.1 1 10 100 1000 10000 100000 1000000 10000000 Ma x i m u m C o n c e n t r a t i o n ( m g / L ) Sutton Weatherspoon Lee Cape Fear Roxboro Mayo Asheville 0.001 0.01 0.1 1 10 100 1000 Mi n i m u m C o n c e n t r a t i o n ( m g / L ) Sutton Weatherspoon Lee Cape Fear Roxboro Mayo Asheville Page 14 Table 2.4: Constituents to hold constant at average concentrations in PHREEQC geochemical model Constituent Molecular Weight Average (µg/L) Average (mol/L) Antimony 121.76 2.28E+00 1.87E-08 Arsenic 74.92 8.46E+01 1.13E-06 Beryllium 9.01 1.94E+01 2.15E-06 Boron 10.81 1.42E+03 1.32E-04 Cadmium 112.41 1.82E+00 1.62E-08 Chromium 52.00 1.22E+01 2.34E-07 Cobalt 58.93 2.72E+01 4.61E-07 Copper 63.55 8.90E+00 1.40E-07 Lead 207.20 6.24E+00 3.01E-08 Mercury 200.59 1.52E-01 7.60E-10 Molybdenum 95.94 4.74E+01 4.94E-07 Nickel 58.69 2.54E+01 4.32E-07 Selenium 78.96 7.78E+00 9.86E-08 Strontium 87.62 6.22E+02 7.10E-06 Thallium 204.38 5.12E-01 2.51E-09 Vanadium 50.94 9.96E+00 1.96E-07 Zinc 65.41 6.68E+01 1.02E-06 Table 2.5: Constituents to vary in concentration to between minimum, average, and maximum ground water concentrations to check for sorption competition in PHREEQC geochemical model Constituent Mol. Weight (g/mol) Minimum (mg/L) Average (mg/L) Maximum (mg/L) Minimum (mol/L) Average (mol/L) Maximum (mol/L) Aluminum 26.98 5.00E-03 1.88E+00 5.74E+01 1.85E-07 6.98E-05 2.13E-03 Barium 137.33 6.00E-03 9.56E-02 1.92E+00 4.37E-08 6.96E-07 1.40E-05 Calcium 40.08 4.40E-02 5.10E+01 5.64E+02 1.10E-06 1.27E-03 1.41E-02 Carbonate Alkalinity 60.01 0.00E+00 1.44E+02 3.80E+02 0.00E+00 2.41E-03 6.33E-03 Chloride 35.45 1.10E+00 3.70E+01 5.70E+02 3.10E-05 1.04E-03 1.61E-02 Iron 55.85 1.00E-02 8.42E+00 2.14E+03 1.79E-07 1.51E-04 3.83E-02 Magnesium 24.31 7.00E-03 1.40E+01 2.81E+02 2.88E-07 5.76E-04 1.16E-02 Manganese 54.94 5.00E-03 1.28E+00 4.55E+01 9.10E-08 2.32E-05 8.28E-04 Nitrate (as N) 14.01 1.00E-02 5.75E-01 2.50E+01 7.14E-07 4.10E-05 1.78E-03 Potassium 39.10 1.23E-01 4.82E+00 1.91E+02 3.15E-06 1.23E-04 4.89E-03 Sodium 22.99 4.52E-01 3.15E+01 5.61E+02 1.97E-05 1.37E-03 2.44E-02 Sulfate 96.06 1.10E-01 1.31E+02 1.80E+04 1.15E-06 1.36E-03 1.87E-01 Sulfide 32.07 1.00E-01 4.29E-01 4.18E+00 3.12E-06 1.34E-05 1.30E-04 Page 15 2.3. Pourbaix diagram modeling To gain an understanding of the aqueous chemical species of each constituent of interest, Pourbaix diagrams were generated using Geochemist Workbench v10. To perform these simulations, the WATEQ4F database was utilized because this is the same database used in PHREEQC modeling of the sorption behavior described below. However, Se and V were not available in the Geochemist Workbench database. Instead, the LLNL.v8.r6+ database was used to generate the Pourbaix diagrams for Se and V described below. Constants for Se and V were added to the PHREEQC database for the sorption modeling below. However, based on the similarity of the revised WATEQ4F database used in PHREEQC modeling and the LLNL.v8.r6+ database, the speciation exhibited in the Pourbaix diagrams below is representative of the species. In these Pourbaix diagrams, the EH and pH measurements from Table 2.3 are shown as individual datapoints. A generic groundwater chemistry containing 500 ppb of each constituent of concern was used in the simulations (Table 2.6). These concentrations are generally higher than the concentrations observed in groundwater samples from the sites considered in this report. Therefore, if precipitation is not observed in these diagrams for the EH-pH regions of interest, it will not be occurring for lower concentrations which would be less saturated. The dominant aqueous species is shown in the blue regions and dominant precipitated solid phases are shown in yellow regions. Table 2.6: Concentrations of reagents used to generate Pourbaix diagrams Species Concentration (ppm) Concentration (mol/L) CaSO4. 2H2O 20.0 1.47 x 10-4 MgSO4 5.0 4.17 x 10-5 Na(HCO3) 10.0 1.19 x 10-4 Arsenic 0.5 6.67 x 10-6 Barium 0.5 3.64 x 10-6 Boron 0.5 4.62 x 10-5 Cobalt 0.5 8.49 x 10-6 Selenium 0.5 6.33 x 10-6 Vanadium 0.5 9.82 x 10-6 Chromium 0.5 9.66 x 10-6 Nitrate 1.5 2.43 x 10-5 Manganese 0.5 9.00 x 10-6 It is important to note in these diagrams that only the most abundant aqueous species is shown in these plots. There are numerous aqueous and mineral species contributing to the reactivity of these systems. These diagrams only serve to show major trends in the speciation. More detailed calculations using PHREEQC consider all aqueous species involved and changes with respect to EH and pH as done in these Pourbaix diagrams. However, in those models sorption is considered and distribution coefficients are calculated which consider all of the chemical species present under a given set of conditions. Thus, while these Pourbaix diagrams are useful tools to identify the major species, it is important to note some limitations: Page 16  The dividing lines between boxes are where species may be equal but there is no information in the diagram regarding the uncertainty of the simulation or the change in speciation as pH and EH moves away from the boundary lines. So there may be significant concentrations of other species present which cannot be seen on the diagrams.  The speciation is also considered only for the conditions given (listed in Table 2.1). Altering the concentrations of aqueous constituents may influence the data.  The Pourbaix diagrams report the activity of species, not molar concentrations. So corrections must be made to get molar units or mass units that are typical measures of concentration.  These Pourbaix diagrams show only the aqueous species and precipitates with no consideration of sorption. Therefore, when comparing these with ground water measurements at the site, some consideration must be made regarding the potential for a species to be present in the subsurface but sorbed to the solid phase and not present in the ground water. A notable example of the significance of this is discussed below with regard to As speciation. The Pourbaix diagrams predict that As(V) will be the dominant oxidation state in many waters. However ground water speciation measurements indicate that As(III) is the dominant aqueous oxidation state. Since As(V) sorbs strongly to mineral surfaces under the pH of the ground waters, the As(V) may indeed be present in the system but sorbed to the mineral surface and not measured in ground water samples. 2.4. The use of Kd values In this report the PHEEQC model, which predicts both aqueous and solid phase speciation based on thermochemical principles, is used to calculate Kd values and examine how pH, EH, and ground water ion concentrations influence the predicted Kd values. The stability constants used in the PHREEQC database to describe chemical reactions are on a log scale. Therefore, small differences in the stability constants can have a large impact on the predicted Kd values. As an example of this phenomenon, a plot of the Kd versus fraction sorbed (assuming a 50 g/L suspension of sorbent) is shown in Figure 2.9. The Kd values were calculated using Equation 2.1. ܭௗ ൌ ሾெሿೞ೚೗೔೏ ሾெሿೌ೜ೠ೐೚ೠೞ ൌ ሾெሿ೅೚೟ೌ೗ିሾெሿೌ೜ೠ೐೚ೠೞ ሾெሿೌ೜ೠ೐೚ೠೞ ∗௏ ௠ Equation 2.1 where [M]solid is the sorbed concentration of a constituent M in units of mol/kgsorbent, [M]aqueous is the aqueous concentration of the constituent in units of mol/L, [M]Total is the total initial concentration of the constituent in units of mol/L, V is the volume of the sample in L, and m is the mass of sorbent in kg. Thus V/m is the inverse of the suspended sorbent concentration (50 g/L in the simulation below). Equation 2.1 can be rearranged to estimate the sorbed fraction of the constituent as: ݂௦௢௥௕௘ௗ ൌ1െ݂௔௤௨௘௢௨௦ ൌ1െ ଵ ଵା௄೏ ೘ ೇ Equation 2.2 The figure is meant to illustrate the fact that at Kd values less than 1 or greater than 1000, only small increase in the concentration of sorbed ions can cause orders of magnitude differences in the predicted Kd values. Such small differences would be difficult to determine experimentally based on analytical equipment resolution or detection limits. Thus, in many cases discussed below, very low or very high Kd values are reached which could not be determined in many laboratory studies. Page 17 Figure 2.9: Theoretical relationship between Kd values and the predicted sorbed fraction within a hypothetical 50 g/L suspension of sorbent. Numerical values are provided to the right to demonstrate the small change in the fraction sorbed with increasing Kd. 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 1.0E-03 1.0E-01 1.0E+01 1.0E+03 1.0E+05 1.0E+07 Fr a c t i o n S o r b e d Kd (L/kg) Page 18 3. GEOCHEMICAL MODELING OF ARSENIC 3.1. Pourbaix diagram analysis Under mildly oxidizing to strongly oxidizing conditions, arsenic can exist as the arsenate (AsO4-3) and arsenite (AsO3-3) oxoanions (Figure 3.1). Both are weak acids and persist in solution as HxAsOyx-2y species [35]. Under the pH an EH conditions expected at each site, both As(III) and As(V) may be present as evidenced by the range of values in the Pourbaix diagram below. Relatively low pH values will favor the protonated forms of As(III) and As(V) as H3AsO3 and H2AsO4-, respectively. As the pH increases and the redox potential decreases, the arsenite species (As(III)) could persist as the neutrally charged H3AsO4 or anionic H2AsO3-. Such changes in redox speciation or protonation state can have profound impacts on the mobility of arsenic. Changes in ionic charge will alter the strength of interactions with mineral surfaces. Generally as the pH decreases and mineral surfaces develop increasingly positive net surface charges, sorption of As(III) and As(V) oxoanions will increase [36, 37]. Reduction of As(V) to As(III) will cause greater overall mobility of As because of the lower sorption affinity of As(III) relative to As(V) [38]. As discussed above, the stronger sorption of As(V) would preferentially remove As from the ground waters measured in Figure 2.5 above and thus comparison of the groundwater speciation and the modeled speciation in this work must also consider the influence of sorption that the model accounts for but is not determined in groundwater samples. Figure 3.1: Pourbaix diagram of arsenic species along with the range of pH and EH values examined in PHREEQC modeling (shown by blue symbols). The error bars represent the standard deviation of the average pH and EH value calculated from the combined measurements at all 7 sites for the average. Page 19 3.2. PHREEQC model analysis Changes in ion concentrations, pH, and EH were found to have a significant impact on As sorption. The calculated Kd values from the three modeled ion concentration conditions and the range of pH and EH values are shown in Figure 3.2. There is a wide variation in the Kd values which is a manifestation of 1) a change in arsenic redox speciation between As(III) and As(V) and 2) competition for sorption sites from other anions such as sulfate. The redox speciation changes can be examined by monitoring the changes in the aqueous phase oxidation states as shown in Figure 3.3. The pH/EH condition 5.1/-20 mV is the only condition where arsenic is predominantly in the aqueous phase as As(III). Under other conditions the As(V) ion is the dominant oxidation state. Similar behavior is observed on the solid phase primarily due to the greater sorption affinity of As(V) as compared with As(III). For example, using the speciation output from PHREEQC, separate Kd diagrams could be generated for As(V) and As(III) as shown in Figure 3.4. The Kd values for As(III) range from 10-1 to 102 L/kg and those or As(V) range from 103-106 L/kg. The low values for As(V) at pH/EH 5.1/-20 mV are somewhat erroneous because As(III) is the dominant oxidation state in both solid and aqueous phase samples. This is demonstrated through analysis of the solid phase speciation in Figure 3.5. The solid phase concentrations are plotted in units of mol/L based on the PHREEQC model output despite the apparent conflict of reporting an “aqueous” concentration of a sorbed species. It is clear that As(V) is the dominant sorbed species under almost all conditions, consistent with the high Kd values for As(V) shown in Figure 3.5. Based on these values, it is clear that the oxidation of As(III) to As(V) will result in greater immobility of arsenic. The potential for changes As speciation are significant because the majority of speciation measurements for arsenic in site groundwater indicated As(III) is the dominant oxidation state. Thus, any remedial activity in which more oxidizing conditions are introduced into the system which causes oxidation of As(III) to As(V) would likely result in a decrease in arsenic mobility at the site. Figure 3.2: Distribution coefficient (Kd) for summed As(III) and As(V) species from PHREEQC model. The impact of changing ion concentrations and changes redox speciation are shown by the decrease or increase in Kd values. 1.00E+00 1.00E+01 1.00E+02 1.00E+03 1.00E+04 1.00E+05 1.00E+06 1.00E+07 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / - 103 5.1 / 372 7.1 / 76 To t a l A s K d ( L / k g ) pH / EH (mV) Total As Kd, Min GW Values Total As Kd, Avg GW Values Total As Kd, Max GW Values Page 20 Figure 3.3: Redox speciation of aqueous As from PHREEQC modeling under minimum (top), average (middle) and maximum (bottom) ion concentrations listed in Table 2.5. Note that the aqueous concentrations of As vary due to the extent of sorption at the pH/EH conditions noted. 1.00E-18 1.00E-17 1.00E-16 1.00E-15 1.00E-14 1.00E-13 1.00E-12 1.00E-11 1.00E-10 1.00E-09 1.00E-08 1.00E-07 1.00E-06 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 Aq u e o u s C o n c e n t r a t i o n ( m o l / L ) pH / EH (mV) Aqueous As(V), Min GW Values Aqueous As(III), Min GW Values 1.00E-11 1.00E-10 1.00E-09 1.00E-08 1.00E-07 1.00E-06 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 Aq u e o u s C o n c e n t r a t i o n ( m o l / L ) pH / EH (mV) Aqueous As(V), Avg GW Values Aqueous As(III), Avg GW Values 1.00E-09 1.00E-08 1.00E-07 1.00E-06 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 Aq u e o u s C o n c e n t r a t i o n ( m o l / L ) pH / EH (mV) Aqueous As(V), Max GW Values Aqueous As(III), Max GW Values Page 21 Figure 3.4: Separate distribution coefficients calculated for As(III) and As(V) from PHREEQC model output. To allow easy comparison, the plots are shown on similar y-axes. 1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+00 1.00E+01 1.00E+02 1.00E+03 1.00E+04 1.00E+05 1.00E+06 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 As ( V ) K d (L / k g ) pH / EH (mV) As(V) Kd, Min GW Values As(V) Kd,. Avg GW Values As(V) Kd, Max GW Values 1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+00 1.00E+01 1.00E+02 1.00E+03 1.00E+04 1.00E+05 1.00E+06 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 As ( I I I ) K d (L / k g ) pH / EH (mV) As(III) Kd, Min GW Values As(III) Kd,. Avg GW Values As(III) Kd, Max GW Values Page 22 Figure 3.5: Redox speciation of sorbed As from PHREEQC modeling under minimum (top), average (middle) and maximum (bottom) ion concentrations listed in Table 2.5. 1.00E-18 1.00E-17 1.00E-16 1.00E-15 1.00E-14 1.00E-13 1.00E-12 1.00E-11 1.00E-10 1.00E-09 1.00E-08 1.00E-07 1.00E-06 1.00E-05 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 So r b e d C o n c e n t r a t i o n ( m o l / L ) pH / EH (mV) Sorbed As(V), Min GW Values Sorbed As(III), Min GW Values 1.00E-11 1.00E-10 1.00E-09 1.00E-08 1.00E-07 1.00E-06 1.00E-05 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 So r b e d C o n c e n t r a t i o n ( m o l / L ) pH / EH (mV) Sorbed As(V), Avg GW Values Sorbed As(III), Avg GW Values 1.00E-09 1.00E-08 1.00E-07 1.00E-06 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 So r b e d C o n c e n t r a t i o n ( m o l / L ) pH / EH (mV) Sorbed As(V), Max GW Values Sorbed As(III), Max GW Values Page 23 The Kd values reported in Figure 3.2 are also significantly changed by the concentration of other ions in solution. Note that the total concentration of arsenic in all simulations was 1.13 x 10-6 mol/L as listed in Table 2.4. The model was run in three simulated groundwater conditions containing the minimum, average, and maximum ion concentrations listed in Table 2.5. Generally, the Kd values decrease with increasing concentrations of other ions such as Fe2+ and SO4-2. This is predominantly due to competition for sorption sites between As(V) and As(III) with other anions such as SO42-. As the concentration of SO4-2 in the groundwater simulant increases from 1.15 x 10-6 to 1.87 x 10-1 mol/L based on the values in Table 2.5, the sulfate ion may compete with other anions like arsenate for sorption sites. This behavior can be seen in Figures 3.6 and 3.7 for HFO and HAO, respectively. The dominant surface species in both systems is the protonated surface site ≡FeOH2+ and ≡AlOH2+, due to the relatively low pH of these systems and the relatively high surface protonation constants for HFO and HAO [1,2]. The result of this speciation is that the surfaces generally have a net positive surface charge which can result in greater sorption of anions such as the arsenate, borate, and chromate ions of interest to these sites. The two primary competing ions in these systems are ferrous iron and sulfate. In Figures 3.5 and 3.6, the concentrations of the surface species ≡FeOFe+, ≡Fe-SO4-, ≡AlOFe+, and ≡Al-SO4- all increase as the initial Fe2+ is raised from 1.79 x 10-7 to 3.83 x 10-2 mol/L and the concentration of SO42- is raised from 1.15 x 10-6 to 1.87 x 10-1 mol/L. These increased Fe2+ and SO42- levels cause increased sorption and take up additional sorption sites which otherwise were occupied by trace ions such as the arsenate and arsenite ions examined in this section. There are no indications in the geochemical model that precipitates containing arsenic will form. However, there is some circular logic to this argument because the arsenic concentrations used in the model were based upon measured values in ground waters at the coal ash basin sites. Thus, if the measured arsenic concentrations in those systems were indeed controlled by solubility, the reported values used here would be at or below those levels and would not necessarily indicate a saturated system was present. Therefore, a more accurate comparison is how close potential solid phases are to saturation. A plot of the saturation indices of some relevant phases out of the 300+ possible mineral phases considered in the PHREEQC model is shown in Figure 3.7. The saturation index is a measure of the concentration of an element in solution relative to the maximum possible concentration under equilibrium solubility conditions. Therefore, a saturation index of 1 or greater indicates that the solution is saturated with respect to that ion and will precipitate. A value less than 1 indicates the ion is not saturated and the concentration can be increased before saturation will occur. The values are generally reported in log units. Thus, based on the log saturation indices reported in Figure 3.7, the concentration of arsenic could be increased several orders of magnitude before precipitation would be expected. The mineral scorodite (FeAsO4.H2O) has the highest saturation index of log -2.6 at a pH and EH of 9.1 and -103 mV, respectively. The dominance of scorodite under these conditions is primarily due to the high Fe(II) content facilitated by the low redox conditions (though not sufficiently low to predict reduction of As(III) to As(V)). The mineral mansfieldite (AlAsO4.2H2O) also has high saturation indices and could theoretically precipitate. However, it is noteworthy that both scorodite and mansfieldite commonly form in hydrothermal deposits. Thus, precipitation of these minerals may have significant kinetic limitations. Page 24 Figure 3.6: Distribution of HFO surface site speciation under minimum (top), average (middle), and maximum (bottom) total groundwater ion concentrations from Table 2.5. A shift from the dominance of H+ and OH- dominated surfaces to a mixture of H+, OH-, Fe2+, and SO42- dominated surfaces with increasing dissolved ion concentrations is shown by then changes in relative color distributions. 1.00E-03 1.00E-02 1.00E-01 1.00E+00 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 Fr a c t i o n o f S p e c i e s pH / EH (mV) HFO Surface speciation, minimum GW ion concentrations HFO_OH2+HFO_OH HFO_O-HFO_OFe+HFO_OFeOH HFO_SO4-HFO_OHSO4-2 HFO_OCa+HFO_OMn+HFO_OMg+ 1.00E-03 1.00E-02 1.00E-01 1.00E+00 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 Fr a c t i o n o f S p e c i e s pH / EH (mV) HFO Surface speciation, average GW ion concentrations HFO_OH2+HFO_OH HFO_O-HFO_OFe+HFO_OFeOH HFO_SO4-HFO_OHSO4-2 HFO_OCa+HFO_OMn+HFO_OMg+ 1.00E-03 1.00E-02 1.00E-01 1.00E+00 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 Fr a c t i o n o f S p e c i e s pH / EH (mV) HFO Surface speciation, maximum GW ion concentrations HFO_OH2+HFO_OH HFO_O-HFO_OFe+HFO_OFeOH HFO_SO4-HFO_OHSO4-2 HFO_OCa+HFO_OMn+HFO_OMg+ Page 25 Figure 3.7: Distribution of HAO surface site speciation under minimum (top), average (middle), and maximum (bottom) total ion concentrations from Table 2.5. A shift from the dominance of H+ and OH- dominated surfaces to a mixture of H+, OH-, Fe2+, and SO4-2 dominated surfaces with increasing dissolved ion concentrations is shown by then changes in relative color distributions. 0% 1% 10% 100% 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 Fr a c t i o n o f S p e c i e s pH / EH (mV) HAO Surface specation, average GW ion concentrations HAO_OH2+HAO_OH HAO_O-HAO_OFe+HAO_OHSO4-2 0% 1% 10% 100% 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 Fr a c t i o n o f S p e c i e s pH / EH (mV) HAO Surface specation, average GW ion concentrations HAO_OH2+HAO_OH HAO_O-HAO_OFe+HAO_OHSO4-2 0% 1% 10% 100% 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 Fr a c t i o n o f S p e c i e s pH / EH (mV) HAO Surface specation, maximum GW ion concentrations HAO_OH2+HAO_OH HAO_O-HAO_OFe+HAO_OHSO4-2 Page 26 Figure 3.8: Saturation indices for five relevant As bearing solid phases considered in the PHREEQC model. Other species for which the saturation index never reached a value > -12 are not shown. 3.3. Comparison of measured and calculated As speciation Arsenic redox speciation data is available for several samples from the Sutton, Lee, and Weatherspoon sites. The fraction of As(III) and As(V) measured at each site as a function of EH and pH are shown in Figure 3.9. While pH and EH values are correlated, the data are shown separately for clarity. These samples generally show As(III) as the dominant oxidation state in the ground water. There is no clear indication that the pH or EH of the system has influenced the observed aqueous As speciation and also little difference in the observed speciation at each site. This is not in agreement with the PHREEQC model which predicts As(V) as the dominant aqueous oxidation state under the majority of EH and pH conditions considered. There are three potential reasons for this discrepancy. First the equilibrium constants for the aqueous and sorbed species could be incorrect. While this is always a possibility the constants used in this model are from reliable databases and data compilations and are considered the best available values. The second possibility is that there are redox couples controlling the arsenic speciation which have not been considered in the model. This is also a possibility but all available redox active species have been included in the model based on field measurements. The final and most plausible reason for the discrepancy is that arsenic is not present under equilibrium conditions. A fundamental assumption of the PHREEQC model is that all chemical reactions are at equilibrium. Thus, the modeled speciation is predicting equilibrium conditions. -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 pH 5.6, EH -20 pH 6.5, EH 220 pH 6.9, EH 513 pH 9.1, EH -103 pH 5.1, EH 372 pH 7.1, EH 76 Sa t u r a t i o n I n d e x Scorodite (FeAsO4.2H2O) AlAsO4:2H2O Mn3(AsO4)2:8H2O Ca3(AsO4)2:4w Arsenolite (As2O3) Page 27 Figure 3.9: Arsenic redox speciation measured in aqueous samples from Weatherspoon, Lee, and Sutton sites as a function of pH (top) and EH (bottom). To evaluate the potential for redox disequilibrium in the field samples, the expected redox potential for each groundwater sample was calculated based on the equation: where [e] is the electron concentration in the system which is more commonly noted as pe (pe = -log[e]). This reaction has been written in terms of Fe++ oxidation to Fe(OH)2+ which are the expected aqueous species at the pH under consideration. The value log -18.78 is the value of the equilibrium constant for this reaction. Iron speciation is used for this calculation because the dissolved Fe concentrations are relatively high compared to As and it is likely that Fe would be a significant redox buffer in these 0.0 0.2 0.4 0.6 0.8 1.0 1.2 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 Fr a c t i o n o f A q u e o u s A s S p e c i e s pH Weatherspoon, Fraction As(III) Weatherspoon, Fraction As(V) Lee, Fraction As(III) Lee, Fraction As(V) Sutton, Fraction As(III) Sutton, Fraction As(V) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 -200 -100 0 100 200 300 400 Fr a c t i o n o f A q u e o u s A s S p e c i e s EH (mV) Weatherspoon, Fraction As(III) Weatherspoon, Fraction As(V) Lee, Fraction As(III) Lee, Fraction As(V) Sutton, Fraction As(III) Sutton, Fraction As(V) ሾܨ݁ሺܱܪሻଶା ሿሾܪା ሿଶሾ݁ሿ ሾܨ݁ ାାሿ ൌlogെ18.78 Page 28 systems. The pe can be converted to EH by multiplying by 16.9 V-1. Taking the log of the above equation, the expected pe value based on the ratio of Fe(II) to Fe(III) can be calculated as: The calculated and measured EH values are shown in Figure 3.10. From these data, it is clear that the expected EH values based on the Fe redox couple are higher than the measured values. Thus, Fe is either 1) not the dominant redox buffer in this system causing the measured EH values or 2) Fe speciation is not present under equilibrium conditions. Figure 3.10: Estimated EH values based on the Fe(II)/Fe(III) redox couple compared with the measured values in groundwater samples from Lee, Weatherspoon, and Sutton sites. The solid black line represents perfect agreement between the measured and estimated values. A similar analysis can be done to evaluate the expected As speciation. In this case rather than predicting the EH of the water, the ratio of As(V) to As(III) was predicted using the measured pH and EH values based on the equation: which can be transformed to the following by taking the log form: 0 100 200 300 400 500 600 700 800 0 200 400 600 800 1000 Me a s u r e d E H (m V ) Estimated EH(mV) based on Fe(III)/Fe(II) Couple Weatherspoon HF Lee Sutton ݈݋݃ ሾܨ݁ሺܱܪሻଶା ሿ ሾܨ݁ ାା ሿ െ2݌ܪ൅18.78ൌ ݌݁ ሾܪଶ ܣݏܱସି ሿሾܪା ሿଷ ሾ݁ሿଶ ሾܪଷ ܣݏܱଷ ሿሾܪଶܱሿ ൌ logെ21.197 log ሾܪଶܣݏܱସି ሿ ሾܪଷ ܣݏܱଷሿ ൌ 3݌ܪ ൅ 2݌݁ െ 21.197 Page 29 Using this equation, As(V) is predicted as the dominant oxidation state under all pH and EH conditions for which data are available, consistent with the PHREEQC model values. While this disagreement between the observed and predicted As redox speciation is concerning, the model remains conservative. If the systems approach equilibrium as expected, oxidation of As(III) to As(V) is expected based on the PHREEQC model. Since As(V) sorbs much stronger than As(III), this would result in a decreased mobility of As and attenuation of As in the subsurface. It is also noteworthy that the predicted Kd values for As from PHREEQC are significantly higher than the values required to describe the observed field data using reactive transport modeling (discussed in section 3.4 below). Thus the lower Kd values required for the model are consistent with the dominance of the more mobile As(III) species in groundwater measurements. 3.4. Comparison modeled and experimental Kd values for arsenic The range of Kd values determined from PHREEQC modeling is significantly broader than the range of values measured under laboratory conditions or used in reactive transport modeling. A comparison of these values is shown in Table 3.1. The low values used in the reactive transport model could be an indication of the predominance of As(III) measured in ground water samples. Based on the measured EH values in the ground waters, As(V) is thermodynamically predicted to be the dominant species in most site ground waters. However, redox disequilibrium in natural systems is common and discrepancies between measured redox conditions using platinum electrodes and values calculated based on measured redox speciation are known to exist [12]. Therefore, the PHREEQC model can be considered a conservative estimate. Any remedial action taken in the coal ash basins to promote more oxidizing conditions or simply the allowance of time to approach a redox equilibrium would result in oxidation of As(III) to As(V) which would reduce the overall mobility of arsenic in these systems. Table 3.1 Arsenic Kd values used in reactive transport modeling, measured in the laboratory, and modeled using PHREEQC Site Reactive transport modeling derived Kd value (L/kg)[3-9] Mean Kd value measured by UNCC batch experiments (L/kg) [10-16] Range of values from PHREEQC geochemical model (L/kg) Sutton 9 48 Total As: 15 to 4.5 x 106 As(V): 11.6 to 4.5 x 106 As(III): .03 to 680 Value for Average GW conditions: As(III): 363; As(V) 2.24 x 104 Lee 14 175 Weatherspoon 45 96.9 Roxboro 15 36.6 Asheville 0.1 1242 Mayo 0.12 83 Cape Fear 0 341.8 Page 30 4. GEOCHEMICAL MODELING OF BORON 4.1. Pourbaix diagram analysis As shown in Figure 4.1, boron exhibits relatively simple chemistry existing as either neutrally charged boric acid, noted in the literature as either B(OH)3 or H3BO3, or as a borate anion H2BO3- (also noted as BO2-) which persists above pH 9. Borate exhibits no redox reactions and solely exists as B(III). The relatively simple aqueous speciation of borate is due to lack of affinity to form complexes with other ions. This lack of chemical reactivity also limits borate sorption to mineral surfaces. Thus boron behaves as a highly mobile ion in the subsurface. Figure 4.1: Pourbaix diagram of boron species along with the range of pH and EH values examined in PHREEQC modeling (shown by blue symbols). The error bars represent the standard deviation of the average pH and EH value calculated from the combined measurements at all 7 sites for the average. 4.2. PHREEQC speciation analysis The PHREEQC model predicts relatively low sorption of boron as expected based on the observed mobility of boron in field samples (Figure 4.2). There is relatively little change in the predicted Kd values as a function of pH for the “minimum” ground water containing relatively low concentrations of major ions. This limited influence of pH is consistent with the persistence of boron as the neutral H3BO3 species as shown in Figure 4.1 and the relatively low competition with other major ions for sorption sites. In the models of the average and maximum major ion concentrations from Table 2.5, competition for sorption sites by other major ions results in a decrease in the observed Kd values. The nature of competition is similar to the decrease in arsenic sorption discussed above regarding figures 3.6 Page 31 and 3.7. There are no changes in boron aqueous speciation across this pH range. Furthermore, there are no precipitates containing boron expected to form. The saturation indices for the four boron bearing minerals considered in the model are shown in Figure 4.3 and all are well below 0. Therefore, the changes in Kd shown in Figure 4.2 are only due to changes in pH and the influence of competing ions. As discussed in section 2, sorption was modeled assuming aluminum and iron hydroxide minerals were the dominant sorbing surfaces. Sorption of boron to aluminum hydroxides was predicted to be significantly higher than iron oxides as shown in Figure 4.4. This is consistent with the higher sorption constant for boron interactions with aluminum hydroxides relative to iron oxides [1, 2] and may also indicate the potential for greater sorption to solids containing higher extractable aluminum concentrations. Figure 4.2: Predicted boron Kd values from PHREEQC modeling using the range of ground water (GW) concentrations listed in Tables 2.4 and 2.5. Figure 4.3: Saturation indices for four relevant boron bearing solid phases considered in the PHREEQC model. Other species for which the saturation index never reached a value > -30 are not shown 1.00E-05 1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+00 4.0 / 482 5.6 / -20 6.5 / 2206.9 / 513 9.1 / - 103 5.1 / 372 7.1 / 76 To t a l B K d ( L / k g ) pH / EH (mV) Total B Kd, Min GW Values Total B Kd, Avg GW Values Total B Kd, Max GW Values -30 -25 -20 -15 -10 -5 0 Sa t u r a t i o n I n d e x Pb(BO2)2 Zn(BO2)2 Cd(BO2)2 Co(BO2)2 Page 32 Figure 4.4: Predicted boron speciation from PHREEQC modeling showing aqueous species and sorption to aluminum and iron hydroxides. Data are from model output using the minimum set of groundwater concentrations listed in Tables 2.4 and 2.5. “Al” and “Fe” in the legend represent the HAO and HFO 4.3. Comparison between modeled and experimental Kd values for boron The Kd values predicted from the PHREEQC model along with experimentally measured batch values and values used in reactive transport modeling are shown in Table 4.1. There is some discrepancy between the highest value predicted by PHREEQC (0.34 L/kg) and the highest values used in the reactive transport modeling and measured by batch sorption (4 L/kg). An earlier version of the PHREEQC model which only assumed sorption to iron oxides significantly under predicted the boron Kd with values near 1 x 10-3 L/kg. The stronger sorption of boron to aluminum bearing solids and the inclusion of HAO sorption reactions in the current model are responsible for the higher Kd values predicted by this updated PHREEQC model. It is noteworthy that in the batch sorption experiments either leaching of boron into the aqueous phase (i.e. the solid selected contained native boron which was desorbing) or demonstrated a non-linear sorption isotherm with minimal sorption. Therefore, the low Kd values are expected. Considering the assumptions in the PHREEQC model regarding the sorption site density, background ion concentrations, and a somewhat arbitrary solid phase concentration assumed in the model, the PHREEQC model is generally in good agreement with the other values. The sorption constants could be revised to provide specific Kd values but this would mainly be a “fitting” exercise. This model does not consider alternative reactions for sequestration of boron such as isomorphic substitution into mica[39]. However, the rates of isomorphic substitution are not known and there is no field data to demonstrate if such a process is occurring at these sites. Therefore, this mechanism of substitution is not included in the model. 1.00E-08 1.00E-07 1.00E-06 1.00E-05 1.00E-04 1.00E-03 45678910 Sp e c i e s C o n c e n t r a t i o n ( m o l / L ) pH Aqueous H2BO3- HAO-H2BO3 HFO-H2BO3 Page 33 Table 4.2: Boron Kd values used in reactive transport modeling, measured in the laboratory, and modeled using PHREEQC. Units in each column are L/kg. Site: Reactive transport modeling derived Kd value (L/kg)[3-9] Mean Kd value measured by UNCC batch experiments (L/kg) [10-16] Range of Kd values from PHREEQC geochemical model Sutton 0 1.7 Range: 1.1 x 10-5 to 0.34 Geometric mean: 8.0 x 10-3 Value for average GW conditions: 0.013 Lee 0 and 3.5 4 Weatherspoon 1 to 4 2 Roxboro 1 0 Asheville 0.1 2.7 Mayo 0.12 0 Cape Fear 1 0 Page 34 5. GEOCHEMICAL MODELING of CHROMIUM 5.1. Pourbaix diagram analysis Under the range of EH-pH conditions representative of the seven sites, chromium is dominated by the trivalent oxidation state (Cr(III), Figure 5.1) with only one EH-pH value falling within the zone of Cr(VI) stability. The trivalent state exists as Cr3+ at low pH and undergoes hydrolysis to form cationic CrOH+2 and Cr(OH)2+, neutrally charged Cr(OH)3(aq), and anionic Cr(OH)4- species with increasing pH. These hydrolysis reactions resulting in cationic species have the potential to sorb to mineral surfaces with increasing pH and/or form discrete precipitates (i.e. Cr2O3) provided the concentration of Cr is sufficiently high. The hexavalent phase exists as the anions HCrO4- and CrO4-2 at environmentally relevant pH values. These are generally soluble states but exhibit moderately strong sorption affinity to metal oxide minerals such as iron oxides [1, 38] which is strong at low pH and decreases with increasing pH as the mineral surface develops an increasingly negative charge. Figure 5.1: Pourbaix diagram of boron species along with the range of pH and EH values examined in PHREEQC modeling (shown by blue symbols). 5.2. PHREEQC speciation analysis Plots of the Kd values from the PHREEQC model as a function of the pH and EH are shown in Figure 5.2. The pH and EH values used in the model are from the range provided in Table 2.3. Under the range of redox conditions examined, there is almost no dependence on the reduction potential (EH) of these systems. The lack of dependence on EH is because trivalent Cr(III) is the dominant oxidation state for the majority of systems examined. Page 35 This is consistent with the dominance of Cr(III) in field samples from the Weatherspoon site presented in Table 5.1. Data for other sites is either not available or the concentrations of chromium are below detection limits required for speciation analysis. Therefore, Kd values calculated for the total chromium in the system are almost exactly the same as those calculated for only the Cr(III) fraction. (Figure 5.2). The only model output containing a significant fraction of HCrO4- or CrO4- is the relatively high EH system (pH 6.9, EH 513 mV). However, due to the strong sorption of Cr(VI), the predicted Kd values remain high. The prevalence of Cr(III) can be seen from the plot in Figure 5.3 comparing the total aqueous chromium and the fraction of Cr(III). Due to the dominance of Cr(III) across almost all EH values under consideration, the sorption of chromium can be more clearly demonstrated by considering the profound influence of pH. As shown in Figure 5.4, the sorption of chromium increases by several orders of magnitude as the pH increases. This behavior is a characteristic of a positively charged cation (in this case the Cr(III) species CrOH+2, and Cr(OH)2+) forming stronger surface complexes as the pH increases and the mineral surfaces develop an increasingly negative surface charge. While Cr(III) remains the dominant oxidation state of Cr at high pH values, sorption decreases as the concentrations of other groundwater ions (such as SO4-2) increase in the groundwater simulants listed in Tables 2.4 and 2.5. This decrease in chromium sorption is shown by the increasing aqueous phase concentrations in Figure 5.3 and the decrease in Kd values from Figure 5.2. The predicted Kd values for chromium sorption exhibit a strong dependence on pH and a relatively small dependence on other ion concentrations. There is relatively little difference between the sorption of chromium under the three groundwater concentrations examined, which is primarily due to the strong sorption of chromium under all conditions to the point that other ions are not capable of outcompeting chromium for sorption sites. The notable exceptions to this observation are the decreased sorption of chromium at low pH values (4.0 and 5.1) and pH 9.1 under the maximum ion concentration system and the general decrease in sorption under the “maximum” groundwater ion concentrations. The decreased sorption at pH 4.0 and 5.1 appear to be due to increased formation of aqueous CrSO4+ and CrHSO4+2 aqueous species (shown in Figure 5.5). These chromium sulfate complexes do not sorb and thus essentially compete with the mineral surfaces for Cr. As the total groundwater sulfate concentrations increase (Table 2.5), the concentration of CrSO4+ and CrHSO4+2 complexes increases and the extent of Cr sorption decreases. As the pH increases, formation of aqueous chromium sulfate complexes becomes less favorable. Instead neutrally charged Cr(OH)3(aq) and anionic Cr(OH)4- become the dominant aqueous species. There is little difference in the fraction of these species at pH 9.1 with the changing groundwater ion concentrations going from minimum to maximum values listed in Table 2.5 (shown in Figure 5.5). Therefore, the decrease in the Kd values at pH 9.1 reported in Figure 5.2 is not due to a change in the aqueous chromium speciation. Rather, this decrease is likely due to competition for mineral surface sorption sites from Fe2+ and other ions that occurs at pH 9.1 as demonstrated in Figures 3.6 and 3.7. Table 5.1: Chromium redox speciation determined in Weatherspoon water samples [17-23]. pH EH Fraction Cr(VI) Fraction Cr(III) 3.6 531.1 0.002 0.998 5.8 129.0 0.015 0.985 5.5 97.0 0.016 0.984 5.0 235.0 0.013 0.987 6.2 99.8 0.014 0.986 5.3 119.0 0.013 0.987 Page 36 Figure 5.2: Predicted chromium Kd values from PHREEQC modeling using the range of ground water (GW) concentrations listed in Tables 2.4 and 2.5. The Kd values of the total chromium fraction (Cr(III) + Cr(VI)) are shown in the top panel and the bottom panel shows only the Cr(III) Kd values. The similarity between the two graphs indicates the Kd values for Cr(VI) are relatively small (average of 36 L/kg). 1.00E+00 1.00E+01 1.00E+02 1.00E+03 1.00E+04 1.00E+05 1.00E+06 1.00E+07 1.00E+08 1.00E+09 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 To t a l C r K d (L / k g ) pH / EH (mV) Total Cr Kd, Min GW Values Total Cr Kd, Avg GW Values Total Cr Kd, Max GW Values 1.00E+00 1.00E+01 1.00E+02 1.00E+03 1.00E+04 1.00E+05 1.00E+06 1.00E+07 1.00E+08 1.00E+09 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 Cr ( I I I ) K d (L / k g ) pH / EH (mV) Cr(III) Kd, Min GW Values Cr(III) Kd, Avg GW Values Cr(III) Kd, Max GW Values Page 37 Figure 5.3: Predicted concentrations of total aqueous chromium and trivalent chromium modeled using the range of ground water (GW) concentrations listed in Tables 2.4 and 2.5. Figure 5.4: Predicted concentrations of total aqueous chromium modeled using the range of ground water (GW) concentrations listed in Tables 2.4 and 2.5. 1.0E-15 1.0E-14 1.0E-13 1.0E-12 1.0E-11 1.0E-10 1.0E-09 1.0E-08 1.0E-07 4.0 / 482 5.6 / -20 6.5 / 220 6.9 / 513 9.1 / -103 5.1 / 372 7.1 / 76 Co n c e n t r a t i o n ( m o l / L ) pH / EH (mV) Total Aqueous Cr, Min GW Values Aqueous Cr(III), Min GW Values Total Aqueous Cr, Avg GW Values Aqueous Cr(III), Avg GW Values Total Aqueous Cr, Max GW Values Aqueous Cr(III), Max GW Values 1.0E+00 1.0E+01 1.0E+02 1.0E+03 1.0E+04 1.0E+05 1.0E+06 1.0E+07 1.0E+08 1.0E+09 345678910 To t a l C r K d (L / k g ) pH Total Cr Kd, Min GW Values Total Cr Kd, Avg GW Values Total Cr Kd, Max GW Values Page 38 Figure 5.5: Predicted chromium aqueous speciation at pH 4.0, 5.6, and 9.1 using the range of ground water (GW) concentrations listed in Tables 2.4 and 2.5. Note, the EH values are not discussed here because >99% of the total chromium exists as Cr(III) under these conditions. 0 0.2 0.4 0.6 0.8 1 1.2 MIN GW, pH 4.0 AVG GW, pH 4.0 MAX GW, pH 4.0 Fr a c t i o n o f A q u e o u s C r S p e c i e s CrOHSO4 CrSO4+ Cr(OH)+2 Cr(OH)2+ Cr+3 0 0.2 0.4 0.6 0.8 1 1.2 MIN GW, pH 5.6 AVG GW, pH 5.6 MAX GW, pH 5.6 Fr a c t i o n o f A q u e o u s C r S p e c i e s Cr(OH)3 CrOHSO4 CrSO4+ Cr(OH)+2 Cr(OH)2+ Cr+3 0 0.2 0.4 0.6 0.8 1 1.2 MIN GW, pH 9.1 AVG GW, pH 9.1 MAX GW, pH 9.1 Fr a c t i o n o f A q u e o u s C r S p e c i e s Cr(OH)4- CrO2- Cr(OH)3 Cr(OH)2+ Page 39 In addition to sorption reactions discussed above, Cr may be removed from solution by precipitation. Based on the groundwater concentrations listed in tables 2.4 and 2.5, only Fe2CrO4 is predicted to have a solubility product greater than one (indicating precipitation is possible). This value is only at relatively high pH values where precipitation will be favored due to charge neutralization of the aqueous species. However, it is noteworthy that the ion concentrations used in this model were determined from aqueous measurements of the ions in pore waters. Thus, unless the systems were supersaturated and precipitation was kinetically hindered, prediction of a saturated system would be unlikely. Therefore, a more useful comparison is to note how close each phase is to reaching saturation. Generally, Cr(OH)3 has a saturation index near -6. Thus, the aqueous chromium concentration could increase several orders of magnitude before additional precipitates may form. Formation of such precipitates is more favored at higher pH values. Figure 5.6: Chromium saturation indices predicted using the maximum ground water ion concentrations listed in Tables 2.4 and 2.5. 5.3. Comparison between modeled and experimental Kd values for chromium A comparison of the PHREEQC model predicted Kd values and Kd values measured from batch sorption experiments are provided in Table 5.2. Experimental values are only available for the sorbents from the Roxboro and Asheville sites. The wide range of modeled values spans the range of experimental values. However, as noted in section 2.4, the strong sorption of any ion (chromium in this case) can result in very high Kd values such as those shown in Table 5.2. The range of values for both Cr(III) and Cr(VI) are indicative of the profound influences of pH on sorption of chromium. Sorption of Cr(III) increases with increasing pH owing to its existence as cationic and neutral species. Conversely, Cr(VI) sorption decreases with increasing pH due to the predominance of anionic CrO42- species. Based on the speciation analysis of Cr from both the PHREEQC model and ground water measurements, trivalent Cr(III) is the primary species expected in the aqueous phase. Thus, the Kd range 23 to 6.9 x 108 is the more appropriate range to consider. While pH is the primary variable of concern, complexing anions in ground water can -20 -15 -10 -5 0 5 10 Sa t u r a t i o n I n d e x FeCr2O4 Cr(OH)3(am) Cr(OH)3 MgCr2O4 Page 40 also have a major impact of the Kd values as demonstrated by the ~103 change in Kd from the three groundwater simulants. Table 5.2: Chromium Kd values (L/kg) from batch laboratory studies and modeled using PHREEQC. NM = not measured. Leaching = leaching or no sorption observed. Site: Mean value measured by UNCC batch experiments [10-16] Range of values from PHREEQC geochemical model Sutton NM Total Cr: 23 to 6.9 x 108, Mean 1.6 x 106 Cr(III): 23 to 6.9 x 108 Mean 1.8 x 106 Cr(VI): 1.3 x 10-3 to 1.9 x 105 Mean 36 Lee NM Weatherspoon NM Roxboro 139, Maximum of 830 Asheville 655, Maximum of 20,490 Mayo Leaching Cape Fear NM Page 41 6. GEOCHEMICAL MODELING of MANGANESE 6.1. Pourbaix diagram analysis Manganese can exist in multiple oxidation states ranging from Mn(II) to Mn(VII). Under the EH and pH conditions of the groundwater at each of the seven sites under consideration, Mn(II) is the dominant oxidation state (Figure 6.1). The selected EH and pH values used for PHREEQC modeling, shown in the Pourbaix diagrams in Figure 6.1) all fall within the region where Mn(II) is dominant. One datapoint lies within the region where the mineral rhodochrosite (MnCO3) is saturated. It is noteworthy that several other Mn mineral phases are possible including relatively common soil minerals such as hausmannite (Mn3O4), manganite (MnOOH), and birnessite (MnO2). The minerals pyrolysite (-MnO2) and bixbyite (Mn,FeO3) were also predicted to form but were removed from consideration in the model due their rarity. While these mineral phases are possible, they preferentially occur at high pH and are not expected under the conditions modeled in Figure 6.1 (based on the concentrations in Table 2.6). Therefore, should the groundwater conditions shift to higher pH values, sequestration of manganese via precipitation could occur. These mineral phases typically form under higher pH regions where hydrolysis of Mn2+ can lead to reduced or neutrally charged species MnOH+ and Mn(OH)20, which can facilitate precipitation. However, hydrolysis of Mn2+ to form MnOH+ is not expected below pH 10. Thus, Mn2+ remains the dominant aqueous ion to be considered in these systems. Page 42 Figure 6.1: Pourbaix diagram of manganese species along with the range of pH and EH values examined in PHREEQC modeling (shown by blue symbols). The error bars represent the standard deviation of the average pH and EH value calculated from the combined measurements at all 7 sites for the average. The top plot neglects precipitation of any solid phases which are shown with yellow shading in the bottom plot. Page 43 6.2. PHREEQC speciation analysis As expected from the Pourbaix diagrams shown in Figure 6.1, the dominant aqueous species of manganese is Mn2+ under the range of groundwater conditions examined in the PHREEQC model. This observation is consistent with measurements of Mn oxidation state speciation in Weatherspoon, Lee, and Sutton groundwater as shown in Figure 6.2. Divalent Mn2+ is a weakly sorbing cation and forms weak complexes with groundwater anions such as Cl-, SO4-2, and CO3-2. Due to the weak complexation affinity of Mn2+, it persists as the free ion Mn2+ despite relatively high concentrations of anions used in the model. Though manganese can also exist as Mn(III), Mn(IV), and Mn(VII), >99.99% of the manganese in the model output is present as Mn(II). Based on the region of stability for other manganese oxidation states in Figure 6.1, only very high pH and EH conditions will facilitate formation of Mn(VII). Aqueous species of Mn(III) and Mn(IV) are not expected because trivalent Mn(III) is unstable in aqueous systems and Mn(IV) is generally insoluble and will precipitate from solution as MnO2(s) minerals birnessite or pyrolusite. Figure 6.3: Fraction of Mn(II) relative to total aqueous Mn in Weatherspoon, Lee, and Sutton groundwater as a function of pH. Due to the prevalence of Mn(II), the system EH has little influence over the sorption behavior of manganese in the model output. Thus, the modeled Kd values are shown in Figure 6.3 as a function of pH to demonstrate the major influence of pH on the system. The Kd values are generally low and, as expected based on the low complexation affinity of Mn(II) discussed above, show little change with respect to the changing groundwater ion concentrations. As the pH increases, the mineral surface changes from a net positive to a net negative charge which causes increasing attraction of the Mn2+ cation with increasing pH (exhibited by the increasing Kd values). Unlike the models discussed above for As, B, and Cr, the Mn concentration in the model varied between the minimum, average, and maximum concentrations observed at the seven sites based on the values listed in Table 2.5. The similarity in Kd values from the minimum and average groundwater concentration systems demonstrates the linearity of a Kd model where a change in the total concentration does not cause a change in the observed distribution coefficient. However, at the maximum groundwater Mn concentration of 45.5 ppm (8.28 x 10-4 mol/L), the surfaces may become saturated with Mn and other sorbing ions which leads to a decrease in the overall Kd value. At the average 0.000 0.200 0.400 0.600 0.800 1.000 345678910 Fr a c t i o n o f M n ( I I ) pH Weatherspoon Lee Sutton Page 44 pH and EH conditions of groundwater from all seven sites (pH 6.47, EH 222 mV as listed in Table 2.3), the Kd values are 0.10, 0.15, and 0.02 L/kg for the minimum, average, and maximum groundwater simulants, respectively. Thus, increasing both the groundwater ion concentrations and total Mn to the maximum values (Table 2.5) results in relatively consistent Kd values with approximately a 5x difference in the range of Kd values. Figure 6.3: Predicted concentrations of total aqueous chromium modeled using the range of groundwater (GW) concentrations listed in Tables 2.4 and 2.5. As expected based on the relatively high Mn aqueous concentrations and the prevalence of mineral phases in the Pourbaix diagrams in Figure 6.1, there are multiple minerals containing Mn which have saturation indices close to 0 (Figure 6.4). The saturation index for rhodochrosite is greater than zero in many simulations indicating the mineral could precipitate from solution. However, rhodocrocite generally occurs in hydrothermal systems and is unexpected to form under these site conditions. The common soil minerals manganite, hausmannite, and birnessite are more likely to dominate these systems [40]. The saturation indices for these minerals are well below zero in many cases with the notable exceptions being the systems with higher pH values (9.1 and 7.1) and the pH 6.9, EH 513 mV system. The minerals hausmannite, birnessite, and manganite have higher saturation indices in the pH 6.9, EH 513 mV system because aqueous Mn(II) can be oxidized to higher oxidation state Mn(III) and Mn(IV) under the higher EH conditions (relative to the other models). Due to the higher concentrations of Mn(III) and Mn(IV) under the high EH condition, the saturation indices of hausmannite, manganite, and birnessite are increased. Increasing the pH could increase the saturation index of these minerals further. Therefore, while these model conditions do not predict the formation of a pure Mn mineral phase, increased Mn groundwater concentrations coupled with increases in pH and EH could result in the formation of hausmannite, manganite, and/or birnessite. While pyrolusite is also shown in Figure 6.3, birnessite is the more common form of MnO2 and thus has been used in the discussion above. Without high EH conditions, Mn(II) will persist and the Mn(OH)2(s) mineral pyrochroite may form. The saturation index increases with increasing pH. Therefore, higher pH values could facilitate formation of pyrochroite. The model at pH 9.1 was the highest considered in this model and yielded a pyrochroite saturation index of -1.23. 1.00E-07 1.00E-06 1.00E-05 1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+00 1.00E+01 345678910 To t a l M n K d v a l u e ( L / k g ) pH Mn Kd, MIN GW Values Mn Kd, AVG GW Values Mn Kd, MAX GW Values Page 45 Therefore, approximately a 10x increase in the aqueous Mn(II) concentration at pH 9.1 could cause saturation of the system with respect to pyrochroite and lead to precipitation. Figure 6.4: Manganese saturation indices predicted using the maximum ground water ion concentrations listed in Tables 2.4 and 2.5. 6.3. Comparison between modeled and experimental Kd values for manganese The Kd values from the PHREEQC model agree well with the data available from batch sorption experiments and reactive transport modeling. The Kd values from the PHREEQC model have a wide range due to the strong influence of pH. However, the PHREEQC Kd value of 0.15 L/kg for the average groundwater conditions (i.e. average pH, EH, and ion concentrations from Tables 2.3 and 2.5) is very similar to the values of 0.1 and 0.01 L/kg used to model the transport of Mn at Roxboro and Mayo, respectively. Table 6.1: Manganese Kd values used in reactive transport modeling, measured in the laboratory, and modeled using PHREEQC. Units in L/kg. The hash symbol indicates either Mn was not included in the reactive transport model or was not measured in batch sorption tests. Site: Reactive transport modeling derived Kd values [3-9] Mean value measured by UNCC batch experiments [10-16] Range of values from PHREEQC geochemical model Sutton - - Range : 5.2 x 10-7 to 5.2 Geometric mean: 7.0 x 10-3 Value for average groundwater conditions: 0.15 Lee 0 - Weatherspoon - - Roxboro 0.01 - Asheville - - Mayo 0.1 - Cape Fear - 0.009 -20 -15 -10 -5 0 5 pH 4.0, EH 482 pH 5.6, EH -20 pH 6.5, EH 220 pH 6.9, EH 513 pH 9.1, EH -103 pH 5.1, EH 372 pH 7.1, EH 76 Sa t u r a t i o n I n d e x MnCO3 (Rhodochrocite) MnOOH (Manganite) MnO2 (Pyrolysite) Mn3O4 (Hausmannite) Mn(OH)2 (Pyrochroite) MnO2 (Birnessite) Page 46 7. SUMMARY Based on the model predicted Kd values and the aqueous speciation underlying the models as well as the observational data from field measurements, a list of potential attenuation mechanisms for several constituents was compiled (Table 7.1). The list includes physical attenuation in the form of flow through a system which will cause dilution and which is expected for all elements. Sorption and precipitation are also considered. Sorption is defined broadly and is proposed to account for any process removing aqueous ions via chemical interactions with a surface. Thus, sorption reactions can include ion exchange, surface complexation, sorption to metal oxides, sorption to metal sulfides, and sorption to organic matter. Precipitation broadly includes both homogenous mineral precipitation as well as co-precipitation. Table 7.1: Listing of primary attenuation mechanisms and general geochemical considerations for several constituents of concern  Constituent Physical attenuation Chemical Precipitation Sorption Arsenic √ √ √ The PHREEQC model predicts As(V) as the dominant oxidation state of arsenic under the field measured EH and pH conditions but As(III) is the dominant species measured in ground waters. The reason for this discrepancy is proposed to be due to 1) increased sorption of As(V) relative to As(III) which would remove all As(V) from the ground water and prevent As(V) measurements in samples and/or 2) a kinetic limitation with respect to the As(III)/As(V) oxidation/reduction reactions which prevents the system from reaching chemical equilibrium. However, the observation of As(III) is consistent with the relatively lower Kd values required in the reactive transport modeling efforts compared with the higher Kd values predicted by PHREEQC. Therefore, the reactive transport model represents a conservative estimate. Due to the stronger sorption of As(V), the tendency of the element to move in the subsurface, will decrease as As(III) becomes oxidized to As(V) and sorbs to mineral surfaces. Additionally, the minerals scorodite (FeAsO4.2H2O) and mansfieldite (AlAsO4.2H2O) are near saturation under some pH and EH conditions examined in this model and measured in the field. Thus these minerals may theoretically form but generally are unlikely mineral phases to form in the shallow subsurface. Boron √ √ Boron exists only in the B(III) oxidation state and generally persists as the neutrally charged chemical species boric acid (H3BO3), which is a weak acid and exhibits minimal sorption to mineral surfaces. As the system pH increases, H3BO3 will deprotonate (i.e. release a H+ ion) to form H2BO3- which also sorbs weakly. Boric acid and H2BO3- are the only two aqueous species of boron predicted to occur in this model. Thus, the PHREEQC predicted Kd values for boron are low (1.1 x 10-5 to 0.34 L/kg). These values are slightly lower but generally consistent with the values chosen for reactive transport modeling and those measured in batch laboratory experiments. Precipitation of any boron containing mineral phases is not expected to occur. Therefore, physical attenuation and sorption are the two primary processes which will control the movement of boron in the subsurface. Page 47 Constituent Physical attenuation Chemical Precipitation Sorption Chromium √ √ √ The PHREEQC model output indicated that Cr(III) is the dominant oxidation state in agreement with the Pourbaix diagram in Figure 5.1. The sorption of Cr(III) is significantly stronger than Cr(VI) because Cr(III) persists as a highly charged cation (Cr3+) which readily sorbs to mineral surfaces as the pH increases from acidic to basic conditions. This behavior is in stark contrast to that of Cr(VI) which persists as a weakly sorbing anion (CrO4-) and decreases sorption from acidic to basic conditions. This high charge density of Cr3+ also causes a propensity to form aqueous complexes with anions such as SO4-2 and Cl- which can influence sorption behavior. For example, formation of CrSO4+ appears to be responsible for a decreased Kd relative to baseline conditions in the PHREEQC model presented in this work. The measured aqueous concentrations in groundwater from the seven sites range from below detection to approximately 100 g/L. This concentration range is similar to what was modeled in PHREEQC and is indicates that formation of mineral phases containing Cr may occur under high pH conditions with relatively high Cr concentrations.  √ √ √  Manganese  Manganese is predominantly present as Mn2+ in the PHREEQC model output which is in agreement with the measurements of Mn(II) in groundwater from all seven sites. Sorption of Mn(II) is generally weak and yields Kd values ranging from 5 x 10-7 to 5 L/kg calculated from the PHREEQC model. The mean Kd value of 7 x 10-3 L/kg and the Kd value of 0.15 L/kg for the average groundwater conditions from the PHREEQC model are in reasonable agreement with the range of 0 to 0.10 L/kg used in reactive transport models [3, 4, 8] and the value of 9 x 10-3 L/kg determined from batch sorption experiments with solids from the Cape Fear site. Analysis of the saturation index of mineral phases containing Mn from the PHREEQC model indicates that several common Mn bearing soil minerals (manganite, hausmannite, and birnessite) are near saturation under high pH and high EH conditions. Therefore, precipitation of Mn mineral phases could occur given sufficiently high Mn concentrations and high pH/EH conditions. Based on the above analysis, it is expected that physical attenuation, sorption, and chemical precipitation could all play a role on controlling the movement of Mn in the subsurface.    Page 48    8. REFERENCES 1. Dzombak, D. A., Morel, F. M. M.: Surface complexation modeling : hydrous ferric oxide1990, New York: Wiley. xvii. 2. Karamalidis, A. K., Dzombak, D. A.: Surface Complexation Modeling: Gibbsite2010, Hoboken, NJ: John Wiley and Sons, Inc. . 3. Brame, S. E., Graziano, R., Falta, R. W., Murdoch, L. C.: Groundwater flow and transport modeling report for the H.F. Lee Energy Complex, Goldsboro, NC, 2015: Greenville, SC. 4. Murdoch, L. C., Brame, S. E., Falta, R. W., Graziano, R.: Groundwater Flow and Transport Modeling Report for Mayo Steam Electric Plant, Roxboro, NC, 2015. 5. Graziano, R., Falta, R. W., Brame, S. E., Murdoch, L. C.: Groundwater Flow and Transport Modeling Report for Cape Fear Steam Electric Plant, Moncure, NC, 2015: Greenville, SC. 6. Falta, R. W., Brame, S. E., Graziano, R., Murdoch, L. C.: Groundwater Flow and Transport Modeling Report for L. V. Sutton Energy Complex, Wilmington, NC, 2015: Greenville, SC. 7. Falta, R. W., Brame, S. E., Graziano, R., Murdoch, L. C.: Groundwater Flow and Transport Modeling Report for the W. H. Weatherspoon Power Plant, Lumberton, NC, 2015: Greenville, SC. 8. Murdoch, L. C., Brame, S. E., Falta, R. W., Graziano, R.: Groundwater Flow and Transport Modeling Report for Roxboro Steam Electric Plant, Semora, NC, 2015: Greenville, SC. 9. Falta, R. W., Brame, S. E., Graziano, R., Murdoch, L. C.: Groundwater Flow and Transport Modeling Report for Asheville Steam Electric Plant, Arden, NC, 2015: Greenville, SC. 10. Langley, W. G., Shubhashini, O.: Soil Sorption Evaluation: Mayo Steam Electric Plant, 2015: Charlotte, NC. 11. Langley, W. G., Shubhashini, O.: Soil Sorption Evaluation: Asheville Steam Electric Plant, 2015: Charlotte, NC. 12. Langley, W. G., Shubhashini, O.: Soil Sorption Evaluation: Cape Fear Steam Station, 2015: Charlotte, NC. 13. Langley, W. G., Shubhashini, O.: Soil Sorption Evaluation: H. F. Lee Steam Station, 2015: Charlotte, NC. 14. Langley, W. G., Shubhashini, O.: Soil Sorption Evaluation: Roxboro Steam Electric Plant, 2015: Charlotte, NC. 15. Langley, W. G., Shubhashini, O.: Soil Sorption Evaluation: L. V. Sutton Energy Complex, 2015: Charlotte, NC. 16. Langley, W. G., Shubhashini, O.: Soil Sorption Evaluation: W. H. Weatherspoon Steam Station, 2015: Charlotte, NC. 17. Synterra: Comprehensive Site Assessment Report – Mayo Steam Electric Plant, Roxboro, NCSeptember 2, 2015: Greenville, SC. 18. Synterra: Comprehensive Site Assessment Report – Cape Fear Steam Electric Plant, Moncure, NC, September 2, 2015: Greenville, SC. 19. Synterra: Comprehensive Site Assessment Report – Roxboro Steam Electric Plant, Semora, NC, September 2, 2015: Greenville, SC. 20. Synterra: Comprehensive Site Assessment Report – Asheville Steam Electric Plant, Arden, NC, August 23, 2015: Greenville, SC. 21. Synterra: Comprehensive Site Assessment Report – W.H. Weatherspoon Power Plant, Lumberton, NC, August 5, 2015: Greenville, SC. Page 49 22. Synterra: Comprehensive Site Assessment Report –L.V. Sutton Electric Plant, Wilmington, NC, August 5, 2015: Greenville, SC. 23. Synterra: Comprehensive Site Assessment Report – H.F. Lee Electric Plant, Goldsboro, NC, August 5, 2015: Greenville, SC. 24. Martell, A. E., Smith, R. K.: Critical Stability Constants, Standard Reference Database 46, Version 6.30, 2001, National Institute of Standards, Gaithersburg, MD. 25. Davis, J. A., Coston, J. A., Kent, D. B., Fuller, C. C.: Application of the surface complexation concept to complex mineral assemblages. Environmental Science and Technology 32, 2820 (1998). 26. Davis, J. A., James, R. O., Leckie, J. O.: Surface ionization and complexation at the oxide-water interface.1. Computation of electrical double layer properties in simple electrolytes. Journal of Colloid and Interface Science 63, 480 (1978). 27. Goldberg, S.: Use of Surface Complexation Models in Soil Chemical Systems, in Advances in Agronomy, D. L. Sparks, Editor 1990?, Academic Press, Inc. p. 233. 28. Powell, B. A.: Analysis of geochemical phenomena controlling mobility of ions from coal ash basins at the Duke Energy Asheville Power Plant, 2015: Clemson, SC. 29. Powell, B. A.: Analysis of geochemical phenomena controlling mobility of ions from coal ash basins at the Duke Energy Cape Fear Steam Electric Plant, 2015: Clemson, SC. 30. Powell, B. A.: Analysis of geochemical phenomena controlling mobility of ions from coal ash basins at the Duke Energy H. F. Lee Energy Complex, 2015: Clemson, SC 31. Powell, B. A.: Analysis of geochemical phenomena controlling mobility of ions from coal ash basins at the Duke Energy Mayo Steam Electric Plant, 2015: Clemson, SC. 32. Powell, B. A.: Analysis of geochemical phenomena controlling mobility of ions from coal ash basins at the Duke Energy Roxboro Steam Electric Plant, 2015: Clemson, SC. 33. Powell, B. A.: Analysis of geochemical phenomena controlling mobility of ions from coal ash basins at the Duke Energy L. V. Sutton Energy Complex, 2015: Clemson, SC. 34. Powell, B. A.: Analysis of geochemical phenomena controlling mobility of ions from coal ash basins at the Duke Energy W. H. Weatherspoon Power Plant, 2015: Clemson, SC. 35. Hem, J. D.: Reactions of metal ions at surfaces of hydrous iron oxide. Geochem. Cosmochim. Acta. 41, 527 (1977). 36. Stachowicz, M., Hiemstra, T., Van Riemsdijk, W. H.: Surface speciation of As(III) and As(V) in relation to charge distribution. Journal of Colloid and Interface Science 302, 62 (2006). 37. Strawn, D., Doner, H., Zavarin, M., Mchugo, S.: Microscale investigation into the geochemistry of arsenic, selenium, and iron in soil developed in pyritic shale materials. Geoderma 108, 237 (2002). 38. Usepa: Monitored Natural Attenuation of Inorganic Contaminants in Ground Water. Volume 2. Assessment for Non-Radionuclides Including Arsenic, Cadmium, Chromium, Copper, Lead, Nickel, Nitrate, Perchlorate, and Selenium.2007, 2007. 39. Sposito, G.: The chemistry of soils1989, Oxford: Oxford University Press. 40. Mckenzie, R. M.: Manganese Oxides and Hydroxides, in Minerals in Soil Environments, 2nd Edition, J. B. DixonWeed, S. B., Editors. 1989, Soil Science Society of America: Madison, WI. p. 439.