Arsenite sorption on troilite (FeS) and pyrite (FeS 2 )

Transcription

1 Pergamon doi: /s (02) Geochimica et Cosmochimica Acta, Vol. 67, No. 5, pp , 2003 Copyright 2003 Elsevier Science Ltd Printed in the USA. All rights reserved /03 $ Arsenite sorption on troilite (FeS) and pyrite (FeS 2 ) BENJAMIN C. BOSTICK and SCOTT FENDORF* Department of Geological and Environmental Sciences, Stanford University, Stanford, CA , USA (Received January 11, 2002; accepted in revised form August 5, 2002) Abstract Arsenic is a toxic metalloid whose mobility and availability are largely controlled by sorption on sulfide minerals in anoxic environments. Accordingly, we investigated reactions of As(III) with iron sulfide (FeS) and pyrite (FeS 2 ) as a function of total arsenic concentration, suspension density, sulfide concentration, ph, and ionic strength. Arsenite partitioned strongly on both FeS and FeS 2 under a range of conditions and conformed to a Langmuir isotherm at low surface coverages; a calculated site density of near 2.6 and 3.7 sites/nm 2 for FeS and FeS 2, respectively, was obtained. Arsenite sorbed most strongly at elevated ph ( 5 to 6). Although solution data suggested the formation of surface precipitates only at elevated solution concentrations, surface precipitates were identified using X-ray absorption spectroscopy (XAS) at all coverages. Sorbed As was coordinated to both sulfur [d(as-s) 2.35 Å] and iron [d(as-fe) 2.40 Å], characteristic of As coordination in arsenopyrite (FeAsS). The absorption edge of sorbed As was also shifted relative to arsenite and orpiment (As 2 S 3 ), revealing As(III) reduction and a complete change in As local structure. Arsenic reduction was accompanied by oxidation of both surface S and Fe(II); the FeAsS-like surface precipitate was also susceptible to oxidation, possibly influencing the stability of As sorbed to sulfide minerals in the environment. Sulfide additions inhibit sorption despite the formation of a sulfide phase, suggesting that precipitation of arsenic sulfide is not occurring. Surface precipitation of As on FeS and FeS 2 supports the observed correlation of arsenic and pyrite and other iron sulfides in anoxic sediments. Copyright 2003 Elsevier Science Ltd 1. INTRODUCTION The availability and migration of arsenic, a toxic metalloid, is controlled largely through sorption processes (Cullen and Reimer, 1989; Korte and Fernando, 1991). Under oxic conditions, As is primarily found as pentavalent arsenate, H x - AsO 4 3-x, which adsorbs strongly to Fe and Al (hydr)oxides (Anderson et al., 1976; Waychunas et al., 1993; Manning and Goldberg, 1996; Fendorf et al., 1997; Hiemstra and Van Riemsdijk, 1999). Surface-bound As is released into solution under slightly reducing conditions through the reductive dissolution of the Fe (hydr)oxides (Cummings et al., 1999; Nickson et al., 2000; Zobrist et al., 2000). Arsenate reduction to highly toxic trivalent arsenite, As(OH) 3, may accompany its release into solution, potentially leading to widespread environmental contamination (Nickson et al., 1998; Acharyya et al., 1999; Langner and Inskeep, 2000). Arsenic concentrations typically decrease under anoxic conditions in ocean sediments (Cutter, 1991; Sullivan and Aller, 1996), freshwater lakes (Aggett and O Brien, 1985; Balistrieri et al., 1994), and rivers (Moore et al., 1988; Johannesson et al., 2000). The uptake of arsenic in anoxic environments is strongly correlated with the formation of iron sulfide minerals including pyrite. Additionally, selective extractions suggest that As is highly pyritized (Huerta-Diaz and Morse, 1992; Cooper and Morse, 1996). However, the contributions of precipitation, coprecipitation, and adsorption mechanisms on sulfide minerals has not been resolved (Belzile et al., 1988; Cooper and Morse, 1998; Holmes, 1999). Arsenic sulfides have also been identified in contaminated lakes (Soma et al., 1994) and wetlands (LaForce et al., 2000). Although the formation of discrete arsenic sulfides, such as orpiment (As 2 S 3 ), realgar (AsS), or arsenopyrite (FeAsS), has been proposed to explain As uptake in anoxic environments, arsenic solubility seldom conforms to that of pure mineral sulfides (Sadiq, 1990; Sadiq, 1997). Thus, the specific mechanism of As retention in anoxic systems is not well understood despite the empirical relationship between the presence of sulfide minerals and arsenic solubility. The surface chemistry of pyrite and other iron sulfide minerals has received considerable attention. Surface termination may lead to the formation of unsaturated surface groups (Rosso et al., 1999) that react with water or dissolved sulfide (Guevremont et al., 1998b). Highly reactive defect sites may also occur at Fe or S surface sites partially stabilized by disproportionation (Guevremont et al., 1997; Uhlig et al., 2001). Disproportionation reactions result in a wide variety of surface species, including Fe(III) sulfides and polysulfides, even on pristine surfaces (Nesbitt and Muir, 1994; Herbert et al., 1998; Nesbitt et al., 1998). The defect structures control oxidation by serving as centers for electron transfer (Guevremont et al., 1997; Guevremont et al., 1998a; Bostick et al., 2000a), and along with disproportionation, may promote cation retention (Bancroft and Hyland, 1990; Xie et al., 1996; Bostick et al., 2000b). For example, the adsorption of Ag on pyrite involves sulfur disproportionation and electron transfer (Hiskey et al., 1987; Scaini et al., 1995). FeS 2 8 Ag 4H 2 O N Fe Ag 2 S 6 Ag 0 SO 4 8 H (1) * Author to whom correspondence should be addressed (fendorf@stanford.edu). 909 Other adsorption reactions may also occur. Surface protons may exchange with cations in solution according to traditional

2 910 B. C. Bostick and S. Fendorf surface complexation or ion exchange (James and Parks, 1975; Kornicker and Morse, 1991). m( S H) Me N ( S) m Me m mh (2) where ( S H) is a surface-bound sulfhydryl group. Metal cations may exchange with insoluble sulfides (with divalent cation Z) through a replacement reaction (Gaudin et al., 1959; Pugh and Tjus, 1987; Park and Huang, 1989). Me ZS N MeS Z (3) Such replacement reactions are favored when the resulting sulfide is less soluble than the initial sulfide mineral; they are thus important for more soluble sulfide minerals such as iron and zinc sulfides (FeS and ZnS). These examples illustrate the rich diversity of cation adsorption reactions that are possible on iron sulfide minerals. Anion adsorption on sulfide minerals has received less attention. Anion sorption to sulfide minerals is typically modeled as an exchange reaction with surface hydroxyl and sulfhydryl groups (Gärd et al., 1995; Balsley et al., 1996; Forsling and Sun, 1997; Balsley et al., 1998). Disproportionation or other redox transformations may also couple with anion sorption on sulfide surfaces; however, the extent of oxidation and reduction during anion sorption is not well documented. Sulfide minerals or dissolved sulfide appear to regulate As levels in anoxic environments. Macroscopic studies of arsenic sorption have been performed on sulfide minerals (Grigorev et al., 1976; Grigorev and Pushkarev, 1986; Zouboulis et al., 1993; Hasany et al., 1999). The mechanism of arsenic retention, however, has not been described. Accordingly, we characterize arsenite sorption on troilite (FeS) and pyrite (FeS 2 ) using both traditional solution-phase methods and spectroscopic techniques. Troilite is used because it is the most stable (and presumably least reactive) form of FeS, and thus its reactivity likely represents a minimum for other iron sulfides (most of which are nonstoichiometric and undergo more complex redox reactions). Pyrite is used because it is the most common sulfide mineral in the environment. Solution-phase data provide information about the variables that control As sorption to sulfide minerals. Spectroscopic techniques, such as X-ray absorption spectroscopy (XAS) and X-ray photoelectron spectroscopy (XPS), were used to identify the structure and oxidation state of sorbed As. In combination, the importance of sulfide phases on As regulation in the environment is established. 2. MATERIAL AND METHODS 2.1. Materials All chemicals used were of analytical grade. Synthetic iron sulfide mineral powders were obtained from Strem Chemical ( 99% purity) and were identified as troilite (FeS) and pyrite (FeS 2 ) by X-ray diffraction. These minerals oxidize readily; therefore, care must be taken to prevent oxidation and preserve, as best as possible, the integrity of the sulfide surface. Consequently, the following procedures were used to prevent oxidation. Minerals were first washed with water, then a 0.01 mol/l sulfide solution (ph 7) to reduce any remaining oxidized species, and then washed again with water to remove residual sulfide and other products. Sulfide treatment created similar surfaces to acid treatment for FeS 2 and was superior to acid for FeS, which dissolved in the strong acid solution. Reactions were performed at room temperature (20 to 23 C) in a mixed H 2 /N 2 atmosphere (10% H 2 ) with a Pt catalyst that maintained O 2 concentrations below 1 ppm. Additionally, minerals were stored under nitrogen and the surfaces were carefully cleaned before use. The purity of the cleaned solids was determined by dissolving a known mass of solids and determining Fe and S concentrations using a Thermo Jarrell Ash IRIS inductively coupled plasma optical emission spectrometer (ICP-OES). The resulting solids had nearly ideal Fe:S stoichiometries (within 2% given instrumental error) and were comprised entirely of Fe and S; therefore, the solids were predominantly unoxidized, although their surfaces may still contain oxidized regions. Surface area was determined with a three-point Brunauer-Emmett- Teller (BET) isotherm using N 2 as the adsorbate. Samples were prepared and loaded for BET measurements entirely in a glovebox to minimize sample oxidation. The BET surface areas of FeS and FeS 2 were 3.2 and 41 m 2 /g, respectively. X-ray photoelectron spectroscopy (XPS) was used to characterize the extent of surface oxidation as described below. Scanning electron microscopy (SEM) coupled with energy-dispersive spectroscopy (EDS) was further used to characterize the solid phases; some fine-grained material was noted in the solids as purchased, but it was largely removed by washing. This fine-grained material contained measurable oxygen by EDS, indicating that it was more oxidized than the larger particles that were retained for use Solution-phase Experiments Several parameters, including concentration of As(III), mineral, and sulfide, ph, and ionic strength, can influence the extent of As(III) sorption on FeS and FeS 2. Therefore, FeS and FeS 2 suspensions (with a suspension density of 1 g FeS or FeS 2 /L) were reacted with arsenite (0 to 250 M As(III) initial concentration) under a wide-range of conditions. Arsenic and sulfide were added using 10 mm stock solutions, As(III) as NaAsO 2, sulfide as Na 2 S 9H 2 O. Adsorption isotherms were performed at ph 4, 7, and 9 by buffering 1 g/l mineral suspensions with mol/l acetate, 4-morpholinepropanesulfonic acid (MOPS), and borate, respectively. Ionic strength was varied between mol/l and 0.5 mol/l with 1.0 mol/l NaCl, and initial total sulfide concentration was set between 0 and 2 mm. The effect of ph on sorption was investigated for FeS 2 and FeS suspensions (1 g/l) containing As(III). The initial suspension ph was adjusted to between 2 and 3 (with 0.1 mol/l HCl and 0.1 mol/l NaOH) and then As(III) was added to create a 50 M As(III) solution. A sample of the suspension was filtered (0.2 m pore size), base was added to increase the ph slightly, and the ph was measured after 0.5 h equilibration at each ph. This procedure was repeated until the ph reached 11. Similar experiments were performed using suspensions in which the initial ph was high ( 11) and titrated to 3 with HCl. Residual solutions were analyzed for total Fe, As, and S using ICP- OES. These measurements typically had precisions of 5% for a 1 ppm solution. The extent of sorption was determined by difference between initial and final As concentrations. Selected residual solids were sealed in Kapton film to prevent oxidation, stored under nitrogen, and analyzed by XAS or XPS within 1doftheir preparation Flow-cell Experiments Flow-cell experiments were performed to study the effect of residence time on As sorption to FeS 2. These experiments employed a small (1 cm in diameter and 0.1 cm thick) FeS 2 -loaded disk through which a solution containing As(III) and sulfide was forced using a peristaltic pump (Fig. 1). This thin disk of FeS 2 powder was supported by a filter with a 0.2 m pore diameter and placed directly into the synchrotron beam. This configuration permitted in situ XAS analysis of adsorbed complexes while preventing sample oxidation. Flow velocities between 0.10 ml min 1 and 1.0 ml min 1, corresponding to solution residence times between 15 and 150 s, were used. Residence times were appreciably shorter than reaction times of As(III) sorbed on FeS 2 in batch experiments and are thus particularly suited for the analysis of reaction intermediates. Arsenic(III) solution concentrations were set at 25 M. Care was taken to avoid orpiment (As 2 S 3 ) saturation; sulfide concentrations were kept at 0.5 M. The ph was maintained at 7 using mol/l MOPS added directly to As(III) solutions. The disk equilibration process was monitored by XAS. Spectra were collected before reaction with arsen-

3 As(III) sorption to iron sulfides 911 Fig. 1. Flow cell design illustrating the placement of the FeS 2 disk in the synchrotron beam. ite solution and consecutively following flow of the As(III) solution. Final spectra were obtained once no changes were apparent in the adsorption edge of these rapid scans. The equilibration time was short; 15 to 30 min was typically sufficient. A small volume of soluble As was present that potentially could influence the observed spectrum of sorbed As(III). Blanks were thus run in the presence of solution before column packing to ensure that solution-phase As did not contribute significantly to the spectrum. Controls were also run on FeS 2 disks before As addition to identify contaminants in the mineral suspensions, as well as the presence of As in the column or column housing before As addition. These blanks were devoid of As edges; the spectra collected in flow cell experiments thus were representative of arsenite sorbed on FeS X-ray Absorption Spectroscopy Arsenic K-edge XAS X-ray absorption spectroscopy was performed at the Stanford Synchrotron Radiation Laboratory on beamlines 4-1, 4-3, or 9-3. The storage ring operated at 3.0 GeV and at currents between 50 and 100 ma. Spectra were taken with a Si(220) double-crystal monochromator with an unfocused beam. Incident and transmitted intensities were measured with 15-cm N 2 -filled ionization chambers. Sample fluorescence was measured with either a Stern-Heald or 13-element Ge detector in combination with a 6- x Gefilter. The beam was detuned approximately 50% to reject higher-order harmonic frequencies and to prevent detector saturation. Spectra were internally calibrated by placement of a sodium arsenate standard between the second and third ionization chamber; its inflection energy (first-derivative maximum) was set at 11,874.0 ev. X-ray absorption spectra were collected from 200 to 1000 ev about the K-edge of As (11,867 ev). At least five spectra were collected for each sample and averaged for analysis, which was conducted with WinXAS software (Ressler, 1997). For X-ray absorption nearedge structure (XANES) analysis, the background was subtracted and the jump height normalized to unity. No smoothing of the raw spectra was done to preserve spectral line-shapes, although derivative spectra were smoothed using a 7-point Savitsky-Golay smoothing procedure to decrease spectral noise. Experimental spectra were compared with those of other common As species, including arsenate (Na 2 HAsO 4 ), scorodite (FeAsO 4 2H 2 O), arsenite (NaHAsO 2 ), realgar (AsS), orpiment (As 2 S 3 ), arsenian pyrite (As-substituted FeS 2 ), and arsenopyrite (FeAsS). Multiple scatter peaks were considered but not identified in any samples. XANES spectroscopy provides information about the oxidation state and coordination environment of the adsorbed As. Although XANES spectroscopy also provides some information about the local structural environment of As, extended X-ray absorption fine-structure (EXAFS) spectroscopy is needed to determine the specific nature of this coordination. The (k) (EXAFS) spectrum was isolated by subtracting the background, normalizing the edge height, and fitting a seven-point cubic spline function that followed the envelope of the decaying spectrum. The spectrum was converted from energy to momentum space (k-space) using an E 0 of 11,885 ev. The (k) spectrum was then weighted by k 3 to amplify the upper k-range and Fourier-transformed without smoothing to produce a radial structure function (RSF) using a k-range of approximately 3 to 10 Å 1. Individual shells were isolated through Fourier filtering for analysis; however, final fits were always performed on raw (unfiltered and unsmoothed) spectra. The element (Z), coordination number (CN), distance (R), and the Debye-Waller factor ( 2 ) for each shell were determined by fitting the experimental spectrum using phase and amplitude functions derived using FEFF 8 (Rehr et al., 1991; Zabinsky et al., 1995; Ankudinov et al., 1998). The accuracy of these phase and amplitude functions was confirmed by comparing fits of orpiment, arsenate, and arsenite with known structures. Both single and multiple scattering paths were considered, although no multiple scattering paths were required for fitting. The E 0 was constrained to the same value for all shells; all other parameters were varied during fitting. The accuracy of the fits was estimated using the 2 statistical parameter, for which smaller values correspond to the best fits. Each fit had a reduced 2 of 4000 for unsmoothed k 3 (k) spectra and approximately 300 using RSFs. This procedure accurately determines interatomic distances (within 0.02 Å); however, considerable errors in coordination environment (30% for the first shell and less accurate for more distant shells) result due to the high correlation of CN and 2. Therefore, the coordination numbers were constrained for As-S and As-Fe shells. The identification of different elements is made based on both differences in the phases and amplitude functions of different atom pairs and local structure (i.e., interatomic distances) Sulfur K-edge XANES Spectroscopy Sulfur K-edge XANES spectra were collected to examine changes in S oxidation state that may accompany As sorption. XANES spectra of cleaned FeS and FeS 2 were collected before reaction with As, and after reaction with 1 mm As(III) at ph 7 for 1 h. Sulfur XANES spectra were collected at the National Synchrotron Light Source (NSLS), Brookhaven National Laboratory, on beamline X19A under ambient (wet, He purged) conditions. Samples were prepared by reacting a 1 g/l suspension of FeS or FeS 2 with 1 mm As(III) in a ph 7 MOPS buffer (0.002M). An As-free, buffered suspension was used as a control. After reaction for 1 h, the solids were filtered and the resulting moist paste sealed in Mylar and analyzed immediately. These spectra were collected in fluorescence mode using a 1-element PIPS detector; consequently, the spectra are not surface sensitive. Therefore, it was necessary to use elevated As concentrations to observe spectral changes that result from As(III) sorption on the mineral surface. Spectra were calibrated using a 20 mm sulfate standard solution ( ev). The identity of other sulfur species was determined by comparison with other S species, including FeS, FeS 2, elemental sulfur, thiosulfate, sulfite, and sulfate. The contributions of each species were determined using linear combinations of S spectra as described elsewhere (Huffman et al., 1991; Vairavamurthy, 1998). These methods have errors of 5%, although the variation in fitting of these spectra was smaller than 2% X-ray Photoelectron Spectroscopy X-ray photoelectron spectroscopy (XPS) was performed on a Surface Science Instruments S-Probe equipped with a monochromatized Al K source (1486 ev). The resulting kinetic energies were between 400 and 1300 ev, corresponding to a mean free path of 10 to 20 monolayers; thus, only limited surface sensitivity was achieved in these experiments. The instrument was operated at a resolution of approximately 0.7 ev and a spot size of 0.5 mm. Both clean solids and selected As-sorbed FeS and FeS 2 samples were analyzed to determine the integrity of the sulfide surface and changes that result from sorption. Samples were prepared by drying under nitrogen, followed by mounting the resulting powders on double-sided carbon tape. Survey scans were used to determine the average composition of the surface. The

4 912 B. C. Bostick and S. Fendorf Table 1. Binding energy shifts for sulfur 2p 3/2 spectra of the cleaned and As(III)- reacted FeS and FeS 2 used in these experiments. The energy is correct within about 0.2 ev; relative intensities have uncertainties of Sample Species a E (ev) E (ev) b intensity Relative Cleaned FeS FeS (bulk) Control, ph 7 Disulfide Polysulfide S 2 O SO Cleaned FeS 2 S(-II)-Fe Control, ph 7 FeS 2 (bulk) Polysulfide As-reacted FeS FeS (bulk) g/g, ph 7 Disulfide Polysulfide S 2 O SO As-reacted FeS 2 S(-II)-Fe g/g, ph 7 FeS 2 (bulk) Polysulfide S 2 O SO a Peaks were identified based on the data of Herbert et al. (1998), Nesbitt et al. (1998), and Uhlig et al. (2001). b The difference in line energy from bulk material. Fig. 2. The sulfur 2p 3/2 spectra of cleaned and As(III)-reacted FeS and FeS 2. thickness of the adventitious carbon coating was estimated to be between 10 and 20 Å using C content and an escape depth of 20 Å. Detailed scans of the S- and Fe-2p spectral regions provided information about the oxidation of the surface, and the formation of several surface products. The binding energies were calibrated using the S 2p 3/2 spectral lines of bulk pyrite and FeS. The energy difference between the S 2p 3/2 ands2p 1/2 lines was fixed at 1.18 ev and their peak area ratio set at 2 to minimize the number of variables used in fitting. Fe spectra were fit with Fe(II)-S and Fe(III)-S multiplets in accordance with convention (Uhlig et al., 2001). Peak widths were not constrained; but they were similar to those reported by others. Peaks were identified by comparison with literature values (Karthe et al., 1993; Herbert et al., 1998; Nesbitt et al., 1998; Uhlig et al., 2001). Errors in peak area were estimated by examining the variation in peak area (Nesbitt et al., 1995, 2000). For sulfur, the estimated error for S 2p spectra was 0.02 for small peaks; large peaks had lower errors because of improved counting statistics. 3. RESULTS AND DISCUSSION 3.1. Solid-phase Characterization The S 2p spectra of cleaned FeS and FeS 2 contain two principal peaks (Fig. 2) at and ev for FeS and and ev for FeS 2 ; the doublets are attributed to the 2p 3/2 and 2p 1/2 electrons of bulk FeS and FeS 2, respectively (Knipe et al., 1995; Herbert et al., 1998; Nesbitt et al., 1998). Prevalence of the bulk FeS and FeS 2 peaks indicates that much of the surface is free of oxidation; however, weak signals attributed to other sulfur species are suggested by the fitting (Table 1). For pyrite, a small spectral feature is found at a shift of 1.1 ev from the bulk pyrite peak, a feature attributed to sulfide (S(-II)-Fe) presumably formed through the stabilization of broken disulfide bonds (Nesbitt et al., 1998; Uhlig et al., 2001). Peaks somewhat larger than the S(-II)-Fe peaks are present at higher bonding energy than bulk pyrite, indicative of oxidation products such as elemental sulfur or polysulfides which, along with sulfide, are commonly formed on pristine pyrite through disproportionation reactions. The excess polysulfide relative to sulfide indicates that surface oxidation has occurred at least to a limited extent. Polysulfide, along with thiosulfate and sulfate, are found on FeS as well. The presence of oxidized sulfur species indicates that the FeS surface is more oxidized than that of pyrite. The Fe 2p spectra also contain evidence of surface oxidation (Fig. 3). The FeS 2 spectrum contains three distinct chemical environments, Fe(II)-S, Fe(III)-S, and Fe(III) oxides (Table 2). The Fe(III) peak is very small, consistent with limited oxidation of FeS 2. The FeS spectrum contains these features and a small feature diagnostic of Fe(III) sulfates. The Fe(II) and Fe(III)-S peaks are multiplets, typically found on both pyrite (Uhlig et al., 2001) and mackinawite (Herbert et al., 1998) surfaces. The Fe(III)-OH and Fe(III)-SO 4 feature, similar in binding energy to FeOOH and jarosite, are often found on oxidized iron sulfides (Karthe et al., 1993). Although Fe(III) sulfates are commonly found on FeS 2 (Karthe et al., 1993), they are not detected or are small features in these spectra, presumably due to cleaning procedures. Both Fe and S were detected in cleaning solutions at concentrations up to 30 M, suggesting that at least some of the surface oxidation products are removed. While these data clearly show that the surfaces of FeS and FeS 2 are somewhat oxidized, oxidation may have occurred at least partially during XPS sample preparation, which involves drying the samples. Limited surface oxidation may influence As(III) sorption, potentially altering the extent of sorption or the redox state of sorbed As.

5 As(III) sorption to iron sulfides 913 Fig. 3. The Fe 2p 3/2 spectra of cleaned and As(III)-reacted FeS and FeS Macroscopic Sorption Characteristics Sorption isotherms provide information about the interactions of the mineral and adsorbed species. Arsenite sorption was strongest at higher ph, with pyrite having similar sorption magnitude at ph 7 and 9, both of which are much greater than sorption at ph 4 (Fig. 4). A similar trend was also observed for As(III) sorbed on FeS, although the sorption affinity increased with ph over the entire ph range examined. At low concentrations, sorption resembled a Langmuir isotherm (Fig. 4), typical of anion adsorption in which a finite number of sites of similar energy react until all of the sites are occupied (the monolayer capacity). Adsorption was enhanced Fig. 4. Sorption isotherms for As(III) on 1 g/l FeS (a) and FeS 2 (b) at varying ph values. An inset (c) of the low concentration region of FeS is also included to show the concentration range in which the Langmuir isotherm is followed. The lines are the Langmuir isotherm fits for the entire data set and have confidence intervals of 0.3 mol/m 2 at As concentrations below 20 M, and up to 5 mol/m 2 at the highest As concentration. Table 2. Binding energies for Fe 2p 3/2 spectra of cleaned and As(III)- reacted FeS and FeS 2 used in these experiments. Energies are accurate within about 0.5 ev; the relative intensities have uncertainties of 0.04 or less. Sample Species a E (ev) E (ev) b Relative intensity Cleaned FeS Fe(II)-S multiplet c Control, ph 7 Fe(III)-S multiplet Fe(III)-OH Fe(III)-SO Cleaned FeS 2 Fe(II)-S multiplet Control, ph 7 Fe(III)-S multiplet Fe(III)-OH As-reacted FeS Fe(II)-S multiplet g/g, ph 7 Fe(III)-S multiplet Fe(III)-OH Fe(III)-SO As-reacted FeS 2 Fe(II)-S multiplet g/g, ph 7 Fe(III)-S multiplet Fe(III)-OH a Peaks were identified based on the data of Karthe et al. (1993), Herbert et al. (1998), Nesbitt et al. (1998), and Uhlig et al. (2001). b The difference in line energy from bulk material. c Fe-S multiplets are composed of a principal peak (at the stated energy) and up to three other peaks of decreasing intensity at higher energy.

6 914 B. C. Bostick and S. Fendorf shifted to lower ph (between 4 and 5.5). Comparable edge positions have been observed previously for FeS 2 (Zouboulis et al., 1993) and are distinctly different from those observed for As(III) on iron (hydr)oxides (Fig. 5). Arsenite sorption on metal (hydr)oxides is typically described by a diffuse ph envelope in which sorption is greatest at circumneutral ph and lowest at high ph (Dzombak and Morel, 1990; Hiemstra and Van Riemsdijk, 1999). The characteristic differences between sorption to these metal sulfides and metal (hydr)oxides implies that sorption occurs through a fundamentally different mechanism than occurs on iron (hydr)oxides. Anion sorption on hydroxides occurs through ligand exchange of surface hydroxyl groups with an anion in solution (Anderson et al., 1976): 2 Fe OH As OH 3 N Fe 2 O 2 As OH 2H 2 O (4) The reaction of multiple surface groups results in the formation of multidentate surface complexes. Similar adsorption mechanisms have been proposed for anions on sulfide minerals. For example, iodide adsorption on sulfide mineral surfaces is thought to involve the exchange of surface hydroxyl groups (Balsley et al., 1996; Balsley et al., 1998): Fig. 5. Sorption envelopes for 55 M As(III) on 1 g/l FeS (a) and FeS 2 (b). The error bars reflect the standard errors for the sigmoidal fits for As sorption (the titration in which the ph was increased). at higher As activities, indicative of adsorbate adsorbate interactions and cannot be described by a Langmuir isotherm (Fig. 4), suggesting development of a three-dimensional structure, such as a polymeric cluster or surface precipitate, that is described by a BET isotherm. Deviation from the Langmuir isotherm is most notable in the case of FeS (Fig. 4), which exhibits linear sorption above the adsorption maximum (above 14 mol/g sorbed As). The monolayer capacity depends on the surface area of the solid and is diagnostic of the surface site density. We estimate the surface site density with the portion of the adsorption isotherms that appear to be unaffected by precipitation (Fig. 4). For FeS 2 at ph 7, the adsorption maximum was 231 mol/g, which corresponds to a surface coverage of 5.6 mol/m 2. The As(III) surface coverage on FeS was similar (4.4 mol/m 2 ) but resulted in less sorption (14 mol/g) due to its lower surface area. The adsorption maxima yield an average site area of 2.6 and 3.7 sites/nm 2 for FeS and FeS 2 respectively, similar to the site densities (1 to 3 sites/nm 2 ) typically observed for arsenite adsorption on iron hydroxides (Dzombak and Morel, 1990; Hiemstra and Van Riemsdijk, 1999). Similarities between site densities suggest that adsorption processes may control As(III) binding at lower As concentrations; however, a three-dimensional structure, such as a polymeric cluster or surface precipitate, may occur at higher concentrations. The As(III) sorption isotherms for FeS and FeS 2 were affected by ph (Fig. 4). The ph envelopes for As(III) on FeS and FeS 2 also indicate that sorption increases with increasing ph (Fig. 5). Minimal As(III) was sorbed on FeS below ph 6, with sorption increasing markedly between ph 6 and 8. Arsenite sorption on FeS 2 exhibited a similar adsorption edge but was Hg OH I N Hg I OH (5) Although there are many possible ligand exchange mechanism proposed for As(III) adsorption, many with different OH (or H ) stoichiometries, such ligand-exchange reactions often evolve hydroxyl groups (or consume H ); consequently, these models (Eqns. 4 and 5) predict that sorption is most pronounced at low ph. In contrast to the trends expected in Eqns. 1 and 2, arsenite sorption on FeS and FeS 2 increases at higher ph (Fig. 5). Thus, a reaction mechanism similar to Eqn. 1 does not appear to describe As(III) sorption on iron sulfides. Electrostatic (outersphere) sorption mechanisms can be disregarded based on their ph dependence (iron sulfide minerals are negatively charged above ph 2 to 3; Bebie et al., 1988) and lack of ionic strength dependence (Figs. 5 and 6). Outer-sphere sorption mechanisms also are inconsistent with the observed hysteresis in sorption and desorption (Fig. 5); the lag in desorption suggests a slow process more compatible with the formation of strong, innersphere complexes. Thus, it appears As(III) sorbs to FeS and FeS 2 through an inner-sphere mechanism distinct from those of surface hydroxyl exchange. Sulfide additions may influence As(III) sorption to FeS 2 by stimulating the precipitation of stable arsenic sulfides such as orpiment. Sulfide additions would initially stimulate the precipitation of arsenic sulfides, thereby lowering solution As concentrations. Although these solid phases may indeed be the most stable thermodynamically, the addition of sulfide to the system inhibits sorption rather than promoting it (Fig. 7). Thus, arsenic sulfides are unlikely to explain As sorption in these experiments. Inhibition could occur through competitive adsorption of sulfide and arsenite anions to reactive surface sites, or by formation of soluble and potentially particle-unreactive arsenic(iii) sulfide complexes (Webster et al., 1990; Helz et al., 1995). Alternatively, sulfide (sorbed or in solution) could be a reaction product in the operative sorption mechanism, thereby

7 As(III) sorption to iron sulfides 915 Fig. 6. Effect of ionic strength on As(III) sorption to 1 g/l FeS and FeS 2 at ph 7. inhibiting sorption directly. Spectral information, discussed below, is needed to better discern these possibilities Arsenic Speciation Spectroscopic techniques such as XAS supplement the information gleaned from solution-phase experiments by providing specific data regarding the oxidation state and local structure of sorbed As(III). Spectra of arsenite reacted with FeS and FeS 2 are similar and have an inflection point near 11,868 ev (Fig. 8), considerably lower than arsenite (11,871 ev). Spectra were comparable over As loadings ranging from well below Fig. 7. Effect of sulfide addition on As(III) sorption to 1 g/l FeS 2 at ph 7. Fig. 8. XANES spectra of As standards and As(III) sorbed on FeS and FeS 2 at ph 7. monolayer coverage to the highest As loadings examined, implying that a similar As complex is formed. The edge position of sorbed As is intermediate between AsS (11,869 ev) and FeAsS (11,867 ev); the shift in edge position implies that As(III) has undergone extensive reduction. Unfortunately, XANES spectroscopy cannot discriminate well between different arsenic sulfide species. The reduction of As(III) is confirmed by EXAFS spectroscopy. The EXAFS data are, unfortunately, complicated by As oxidation within the synchrotron beam (Fig. 9). Nevertheless, oxidation was sufficiently slow that it did not appreciably influence the XANES analysis. Oxidation was marked by the growth of a spectral feature at 11,874.2 ev over successive scans a similar edge position to that of arsenate. After 1 to 2 h, oxidation was complete and little of the residual signal remained. Samples were typically only oxidized to a limited extent before placement in the synchrotron beam. Despite the sample degradation, EXAFS analysis remained useful for determining the coordination environment of As (Table 3). The principal EXAFS feature for As on FeS and FeS 2 at ph 4 is comprised of both As-S and As-Fe shells at 2.4 Å (Fig. 10) bond lengths similar to As-Fe and As-S in arsenopyrite (2.35 and 2.37 Å respectively). However, the second shell As-Fe distances are pronounced for FeAsS and are absent for As(III) sorbed on FeS and FeS 2. Arsenic substituted in pyrite also contains both As-S and As-Fe coordination, but bond lengths are shorter than those observed for As(III) on FeS 2 (Savage et al., 2000). Importantly, the local structure of sorbed As species differs from both orpiment and realgar, both of

8 916 B. C. Bostick and S. Fendorf Fig. 9. X-ray absorption spectra of 40 M As(III) sorbed to 1 g/l FeS 2 at ph 7 (38 mol/g As) illustrating the sample oxidation that proceeds during analysis. The XANES spectra (a) show the first seven scans during analysis, and the EXAPS spectra (b) and their respective RSFs show the initial and seventh scan collected. Each scan takes approximately 0.5 h to complete. which lack As-Fe coordination and have shorter As-S bond distances than are observed here. The lack of second-shell features suggests the presence of structural disorder (i.e., the formation of amorphous FeAsS). Structural disorder is common for surface precipitates and is appreciable for other arsenic minerals, including freshly precipitated As 2 S 3 (Helz et al., 1995). EXAFS spectra of sorbed As in ph 7 samples show more evidence of oxidation than ph 4 samples (Fig. 10). A single As-O shell near 1.70 Å, characteristic of arsenate, dominates the spectra; however, a significant second shell characteristic of As-Fe near 2.4 Å is also present at ph 7. Although an As-S shell is presumably also present in ph 7 samples (both FeAsS and arsenian pyrite have first shell As-S distances in this range and are the only common minerals with first-shell Fe-As interactions), spectra were of insufficient quality to resolve As-S and As-Fe features. Minor As-Fe or As-As spectral features are also observed at 3.35 Å, presumably from a combination of the pristine and oxidized surface species. Sulfide was added to poise the suspension at more reducing Fig. 10. The k-weighted (k) spectra (a) and corresponding RSFs (b) of As(III) sorbed on FeS and FeS 2 at ph 4 and 7. The low coverage of samples at ph 4 impacts data quality, while samples at ph 7 show a strong As-O shell indicative of oxidation. The grayed region of the RSF denotes the long As-S shell characteristic of arsenopyrite. conditions. It was, however, inefficient at preserving reduced As species except at concentrations above 0.5 mm, where the solution is oversaturated with respect to orpiment. Indeed, the EXAFS spectra of these sulfide-rich suspensions confirm the presence of As 2 S 3 (Fig. 11). The spectra contain an As-S shell at 2.23 Å and weak As-As shells between 3.2 and 3.5 Å, consistent with the spectra of amorphous As 2 S 3 (Helz et al., 1995) Chemical Intermediates Batch reactions contain evidence for the conversion of arsenite to an amorphous FeAsS phase. Flow experiments offer the potential to examine the precursors to this amorphous solid. Additionally, the short residence times of As in flow experiments limit As oxidation, thus enhancing spectral analysis. The EXAFS spectra of As sorbed on FeS 2 in column experiments appear in Figure 12. Generally, these spectra contained As-O and As-S coordination in the first shell and the clear presence of an As-Fe second shell. Similar increases were noted for As(III) sorbed on FeS 2 at ph 6 and 9. The As-O distance was somewhat shorter than for arsenite in solution (1.73 Å compared with 1.77 Å for arsenite). Although much of this differ-

9 As(III) sorption to iron sulfides 917 Fig. 11. EXAFS and corresponding FT spectra of 25 M As(III) sorbed to 1 g/l FeS 2 at ph 7 poised with 0.5 mm and 1.0 mm sulfide. The first shell is a As-S shell at 2.23 Å, and the more distant features are As-As shells, each characteristic of orpiment. ence is within experimental error, it suggests that the As(III) coordination shell is somewhat altered. Small As-Fe and As-S shells are also noted in the spectra. Arsenite coordination is initially conserved in these intermediate surface complexes; however, the As-S coordination number increases with increasing residence time (decreasing flow rates). The As-Fe shell at 2.81 to 2.85 Å is consistent with arsenite adsorption in a mononuclear, bidentate complex on pyrite. An increase in the intensity of the As-Fe shell (at 2.81 Å) occurs concurrent with the increase in As-S shell intensity, suggesting that increased reaction times increases the fraction of adsorbed arsenite. Thus, increased residence times leads to formation of both the bidentate arsenite complex noted above and the arsenic sulfide complex measured under bulk conditions. In contrast to spectra of batch samples, first-shell As-Fe bonds (at distances of 2.3 to 2.4 Å) are not observed in these samples. Thus, it appears that adsorbed arsenite forms as a precursor to As incorporation into a solid phase similar to that identified in batch experiments. It should be mentioned that the surface species identified in batch experiments might also be a metastable product that would eventually give way to the formation of more stable species, crystalline surface precipitates of FeAsS, As 2 S 3,or scorodite. Suspensions not prepared with additional sulfide are particularly sensitive to subsequent reactions. Such suspensions are far from equilibrium, and mineral dissolution may contribute to the formation of arsenic sulfide minerals. Alternatively, the continual reduction of iron and sulfide in anoxic environments may also influence the structure of sorbed As(III) Iron and Sulfur Speciation Arsenic(III) reduction to arsenic sulfide surface species is readily apparent by XAS. The spectra exemplify different structures than observed for solution-phase arsenite the oxy- Fig. 12. EXAFS (a) and corresponding FT (b) spectra of FeS 2 columns reacted with a solution of 50 M As(III) and 1 M sulfide at ph 7 at flow rates of 0.1 ml/min and 1 ml/min. The first shell is a As-O at 1.73 Å, the second shell is a As-S shell at 2.23 Å, and the final shell is a As-Fe shell at 2.81 Å.

10 918 B. C. Bostick and S. Fendorf Fig. 13. S K-edge XANES spectra of 1 g/l FeS and FeS 2 suspensions before (solid lines) and after reaction (dotted lines) with 1 mm As(III) at ph 7. gen coordination of arsenite is replaced by both As-S and As-Fe shells in the first coordination sphere, thereby forming reduced As species. Such reduction must be paired with the oxidation of Fe(II), S(-II), or S(-I). We probed the changes in S and Fe oxidation state using XPS. Comparing S 2p spectra of As-reacted FeS and FeS 2 with cleaned samples, which were prepared under the sample conditions but not reacted with As(III), indicates a loss in the intensity of the S(-II) spectral feature after As sorption (Fig. 2). The fractional area of the S(-II) species decreases from 0.13 on untreated FeS 2 to 0.05 on As-reacted FeS 2. The fractional area of the S(-II) feature on FeS also decreases, from 0.67 to 0.55 after reaction with arsenite (Fig. 2). An increase in the fractional area of the polysulfide surface components is also noted by XPS; this is likely due to the paired oxidation of sulfur species that accompanies As reduction. Iron 2p spectra also contain some evidence of oxidation (Fig. 3). The intensity of both the Fe(III)-S and Fe(III)-OH peaks increases with sorption. The formation of polysulfides was confirmed by S K-edge spectroscopy (Fig. 13). After reaction with As(III), a large fraction of FeS (nearly 30%) converted to FeS 2, indicating sulfide oxidation to disulfide. The change for FeS 2 after sorption is smaller, in part because the spectrum of polysulfides does not differ appreciably from that of pyrite; however, the lack of more oxidized sulfur species suggests that they are not the major products of arsenite sorption on FeS 2. The spectrum of FeAsS is similar to FeS 2 (Mosselmans et al., 1995). Consequently, the fraction of FeS 2 represents the sum of both mineral species. The reaction stoichiometry for FeS, FeS 2, and As(OH) 3 was estimated based on the extent of FeS oxidation and an approximate sorbed As concentration of 630 mol/g. The conversion of 30% of the S to pyrite (and FeAsS) corresponds with the formation of 1126 mol FeS 2 ; thus, the ratio of As sorbed to pyrite and arsenopyrite produced is 1:1. Additionally, 1.5 mol FeS were consumed (1700 mol) per mole FeS 2 (and FeAsS) produced. Subtle changes in the FeS 2 spectrum were also noted after reaction with As(III) but were difficult to quantify because the white-line features of the longer chain polysulfides and pyrite are at similar energies. Nevertheless, the fraction of elemental sulfur rose to 5% after reaction with As(III), confirming that As sorption induced the formation of longer chain iron polysulfides such as FeS 3 or FeS 4, which have been suggested by others based on XPS and electrochemical data (Nesbitt and Muir, 1994; Chadwell et al., 2001). Additional oxidized Fe and S species may have been released into solution during sorption, thereby going undetected by XPS; however, only a small quantity (less than 3 M) of S and even less Fe was released during sorption (probably as oxidized species). The ratio of Fe and S released per mole of As sorbed is 20 or higher, indicating that very little S is released relative to the quantity of As sorbed. We therefore conclude that evolved soluble Fe and S species are not major reaction products of As sorption. We can postulate an equation for the reaction of arsenite with FeS and FeS 2 using information gleaned from analysis of solid-phase and solution reaction products. Arsenite was reduced to a FeAsS-like precipitate on both FeS and FeS 2. The corresponding oxidation of both surface-bound Fe(II) and (di- )sulfide to Fe(III) (hydr)oxide and polysulfide was also detected. Little Fe or sulfate was evolved into solution through reaction with arsenite; any proposed reaction should thus not produce large quantities of dissolved species. For As(III) sorption to FeS, the following reaction scheme is consistent with the observed reaction products and the reaction stoichiometry determined using S K-edge XANES: 3FeS As OH 3 N FeS 2 FeAsS Fe OH 3 (6) The G 0 for Eqn. 6 is favorable, about 42 kj/mol for the conversion of the pyrrhotite polymorph of FeS and forming ferrihydrite. Thermodynamic data were obtained from Stumm and Morgan (1996) except for FeAsS ( G f kj.mol; Kolonin et al., 1988) and for As(OH) 3 ( G f kj/mol; Pokrovski et al., 1996). Authigenic FeS (mackinawite) is less stable, thus, arsenite sorption on mackinawite would be even more favorable. Arsenite sorption on FeS forms FeS 2 ; however, As(III) reacts with FeS 2 to form longer chain polysulfides such as iron tetrasulfide (FeS 4 ). 7FeS 2 2 As OH 3 N 3FeS 4 2FeAsS 2Fe OH 3 (7) It is not possible to determine the Gibbs free energy change for Eqn. 7 because the thermodynamic properties of iron polysulfides are poorly constrained and the precise identity of the polysulfides is unknown; nevertheless, Eqns. 6 and 7 both describe the observed spectral data very well. Reactions 6 and 7 also predict the ph dependence of sorption. For example, Reaction 7 can be divided into steps forming FeAsS and Fe 3 (Reaction 8) and the hydrolysis of Fe 3 (Reaction 9).

11 As(III) sorption to iron sulfides 919 Table 3. Arsenic local structure for As(III) adsorbed to FeS and FeS 2. a As-O As-S As-Fe As-Fe/As Sample CN R (Å) 2 (Å 2 ) CN R (Å) 2 (Å 2 ) CN R (Å) 2 (Å 2 ) CN R (Å) 2 (Å 2 ) 2 Arsenian FeS 2 1 f f f (873) FeAsS 1 f f f,2 f 3.06, , (542) As 2 S 3 3 f f, 2.5 f 3.19, , (342) 13 mol/g As(III) on 1 f f (501) FeS, ph 4 40 mol/g As(III) on 1 f f (537) FeS 2,pH4 6 mol/g As(III) on FeS, f f (566) ph 7 36 mol/g As(III) on f f (978) FeS 2,pH7 41 mol/g As(III) on FeS 2,pH7,0.5mM sulfide (Fig. 11) f f 1 f, 2.5 f 3.21, , (206) 27 mol/g As(III) on FeS 2,pH7,1mM sulfide (Fig. 11) 50 M As(III) on FeS 2 column, ph 9, 0.1 ml min 1 flow rate (Fig. 12) 50 M As(III) on FeS 2 column, ph 9, 1.0 ml min 1 flow rate (Fig. 12) f 1 f, 2.5 f 3.19, , (345) f f f 2143 (679) f f f 2249 (382) f: fixed during fitting. a The spectra for these fits are shown in Figure 10 unless otherwise noted in the sample description. The coordination number (CN) is typically accurate to within 1, the interatomic distance (R) within 0.02 Å; 2 represents the variance in R. For all of the media, E 0 was 11,869 ev. The 2 refers to the 2 for the final fit, the 2 is the improvement in 2 from including the least significant shell. 3FeS As OH 3 N FeS 2 FeAsS Fe 3 3OH (8) Fe 3 3OH N Fe OH 3 (9) Reaction 5, which would be the predominant reaction under acidic conditions, is unfavorable ( G kj/mol). In fact, Reaction 3 only becomes thermodynamically favorable when considering the hydrolysis of the Fe 3 (Reaction 6, G kj/mol). Consequently, arsenite would be expected to adsorb only under conditions sufficiently basic for Fe(III) precipitation. 4. CONCLUSIONS Arsenite sorption on FeS and FeS 2 results in the formation of an FeAsS-like surface precipitate. Both Fe and S must be incorporated into the surface precipitate; thus, the iron sulfide mineral surface is consumed as a consequence. Such a precipitation mechanism influences the retention characteristics of arsenite profoundly. Arsenite sorption is minimal at low ph, in contrast to its behavior on most (hydr)oxide mineral surfaces. The difference in sorption characteristics needs to be considered when evaluating the potential of sulfide minerals to sequester As in natural environments. The high affinity of arsenite for sulfide minerals may regulate As solution concentrations in reducing environments through the formation of FeAsS-like phases in slightly sulfidic solutions and through the formation of As 2 S 3 in highly sulfidic zones. Thus, thermodynamic equilibrium with arsenic sulfide minerals is achieved only in the presence of appreciable sulfide concentrations (Sadiq, 1997). In fact, additions of sulfide below orpiment saturation inhibit As(III) sorption despite the formation, possibly by inhibiting mineral dissolution or through competitive sorption. The association of As with pyrite suggested by selective extraction (Huerta-Diaz and Morse, 1992; Cooper and Morse, 1996) is apparent because FeAsS formation occurs on FeS and FeS 2 mineral surfaces. In both cases, arsenic precipitates as stable, insoluble sulfide phases on the surface of iron sulfide minerals, suggesting that sorption may be an effective and relatively irreversible process that can successfully regulate As concentrations in reducing environments. However, iron, arsenic, or sulfide oxidation may lead to the release of oxidized As species, as process similar to the reductive dissolution of arsenate adsorbed to iron (hydr)oxides. The effect of these oxidation processes needs to be examined to determine the long-term stability of As sequestered by iron sulfide minerals. Table 3. Acknowledgments The authors wish to thank the anonymous reviewers, as well as David Rickard and the associate editor, Frank A.