Kinetics and Products of Chromium(VI) Reduction by Iron(II/III)-Bearing Clay Minerals

Hexavalent chromium is a water-soluble pollutant whose mobility can be controlled by reduction of Cr(VI) to less soluble, environmentally benign Cr(III). Iron(II/III)-bearing clay minerals are widespread potential reductants of Cr(VI), but the kinetics and pathways of Cr(VI) reduction by such clay minerals are poorly understood. We reacted aqueous Cr(VI) with two abiotically reduced clay minerals: an Fe-poor montmorillonite and an Fe-rich nontronite. The effects of ionic strength, pH, total Fe content, and the fraction of reduced structural Fe(II) (Fe(II)/Fe(total)) were examined. The last variable had the largest effect on Cr(VI) reduction kinetics: for both clay minerals, the rate constant of Cr(VI) reduction varies by more than three orders of magnitude with


Introduction
Chromium(VI) is an important groundwater pollutant that also occurs naturally in regions rich in ultramafic rocks. 1,2 Chromium(VI) is generally mobile in ground and surface waters, 3 whereas Cr(III) is an environmentally benign micronutrient 4 with low solubility. 5 Chromium oxidation and reduction have been extensively studied to better understand the behavior of Cr, especially in subsurface environments where Cr may be present in groundwater and drinking water supplies. 6 Although Cr(III) has few naturally occurring oxidants (e.g., Mn(III,IV)-oxides), 7,8 many naturally occurring reductants can reduce Cr(VI) to Cr(III), including aqueous Fe(II) [9][10][11][12] and Fe(II)-bearing solids. [13][14][15][16][17][18][19][20][21][22] Knowledge of both the kinetics and thermodynamic favorability of each potential reductant is critical for assessing the fate of Cr(VI) in complex natural environments.
Factors that may affect reaction kinetics include ionic strength and pH, which may influence Cr(VI) sorption; the Fe content of the clay mineral, which may affect electron transfer pathways; and Fe(II)/Fe(total) (XFe(II)) within the clay mineral. At circumneutral pH, Cr(VI) is primarily reduced by Fe(II) that resides within the clay mineral structure 19 or is sorbed on the clay mineral surface. 31 To facilitate electron transfer, Cr(VI) is expected to sorb to particle edges 18,27,28 or enter into the clay interlayer space, 16 although electron transfer through a mostly redox-inactive tetrahedral layer in the latter case may be difficult. 32,33 Theoretical calculations suggest that electron transfer through the tetrahedral layer should be extremely slow, 34 yet experimental evidence implies that electrons can transfer to Fe(II) sorbed on the basal surface. 35 Intra-layer electron hopping in Fe-rich clay minerals may enlarge the pool of available Fe(II) and enhance the reduction of edge-sorbed Cr(VI), but such hopping may be irrelevant if the clay mineral is Fe-poor or most Cr(VI) is sorbed to the basal surface.
XFe(II) may affect the rate of Cr(VI) reduction because XFe(II) controls the effective standard reduction potential (E°ef f) of clay minerals. [36][37][38] Linear free energy relationships between log of the rate constant k and E°ef f have been constructed for Cr(VI) reduction by aqueous Fe(II) complexes 9 and nitroaromatic reduction by Fe(II/III)-bearing clay minerals, 23 suggesting that such a relationship may describe Cr(VI) reduction by Fe(II/III)-bearing clay minerals. Previous studies of Cr(VI) reduction by Fe(II/III)-bearing clay minerals have only evaluated single values of XFe(II), 16,22 despite the potential for reaction kinetics to vary by orders of magnitude with systematic changes in XFe(II).
Such variation is especially important because imposing reducing conditions during reductive remediation of Cr(VI) may cause large changes in XFe(II). [39][40][41] In this study, we examined the kinetics of Cr(VI) reduction by two Fe (II/III)-bearing 2:1 clay   minerals, an Fe-poor montmorillonite and an Fe-rich nontronite. The montmorillonite contains isolated Fe(O,OH)6 octahedra, hindering intra-layer electron hopping. 42 In the nontronite, Fe(O,OH)6 octahedra comprise the majority of the octahedral layer, permitting intra-layer electron hopping. 32,43 The effects of pH and ionic strength on Cr(VI) reduction were investigated to elucidate the reaction pathway, and each clay mineral was evaluated at several values of XFe(II).
Chromium-and Fe-bearing products were characterized with K-edge X-ray absorption near edge structure (XANES) spectroscopy.

Materials and Methods
The montmorillonite (SWy-2, Ca0. 52Na0.14K0.01(Al3.23Fe0.42Mg0.56)(Si7.89Al0.14)O20(OH)4) 44 and nontronite (NAu-2, M0. 72(Al0.34Fe3.54Mg0.05)(Si7.55Al0.16Fe0.29)O20(OH)4) 45 used in this study were purchased from the Clay Minerals Society (http://clays.org). In their native state, almost all Fe within the clay mineral structure is Fe(III). 16 The clay minerals were sonicated and size fractionated (0.2-2 μm) to minimize impurities (<2 μm) 46,47 but allow easy separation of clay mineral and solution via syringe filtration (>0.2 μm). Size fractionation was verified with scanning electron microscopy and dynamic light scattering, and the only impurity detected by X-ray diffraction was a small amount of quartz associated with both clay minerals. The clay minerals were saturated with Na + through repeated suspension in 1 M NaCl and centrifugation to maintain consistency with their electrochemical characterization by . 38 X-ray fluorescence spectrometry indicates that the montmorillonite contains 2.3 wt. % Fe and the nontronite 23.7 wt. % Fe, consistent with literature values. 16,19,44,47,48 Further manipulations of the two clay minerals were performed in an anaerobic glovebox. Acidwashed glassware and plasticware were allowed to equilibrate with the glovebox atmosphere for at least 24 hours. Aqueous solutions were made with doubly deionized (DDI) water that had been sparged with N2 and allowed to equilibrate for at least 24 hours in the glovebox. Centrifugation was performed outside the glovebox in capped tubes held in aerosol-tight rotor buckets.
Clay minerals were reduced by sodium dithionite following a standard procedure. [49][50][51][52][53][54] In brief, each clay mineral was suspended in a sodium citrate/bicarbonate buffer to minimize clay dissolution. After adding sodium dithionite, the clay mineral was repeatedly washed with DDI water and size fractionated. The extent of reaction was controlled by varying the duration and temperature of reaction or by using only the stoichiometric amount of sodium dithionite. XFe(II) was determined by linear least-squares fitting of Fe K-edge XANES spectra with the spectra of the fully oxidized and fully reduced clay minerals serving as model compounds ( Figure S1). respectively. The clay mineral, buffer, and any salts were allowed to equilibrate overnight before initiating the reaction by adding 10 μM Cr(VI). Aliquots were taken periodically and immediately passed through a 0.2 μm polyethersulfone (PES) syringe filter to arrest the reaction. Chromium(VI) concentrations in the filtered aliquots were determined spectrophotometrically using diphenylcarbazide, 58 and Fe(II) and Fe(III) concentrations were determined spectrophotometrically using ferrozine. 59 Absorbances of gravimetric standards were reproducible within 3%.
Solid reactants and products were characterized using Cr and Fe K-edge XANES spectroscopy, which was carried out at beamlines 4-1 and 11-2 at the Stanford Synchrotron Radiation Lightsource as described in the SI. The initial concentration of Cr(VI) was varied to produce molar Cr(aq):Fe(clay mineral) ratios from 0.00033 to 0.33. XANES scans were calibrated and averaged in SixPack 60 before being normalized using Demeter software 61 by fitting a first-order polynomial to the pre-edge region and a second-order polynomial to the post-edge region. Chromium and Fe speciation were determined by linear least squares fitting to a library of XANES spectra of Cr-and Fe-bearing standards. 62 Scanning transmission X-ray microscopy (STXM) was performed on nontronite reaction products at beamline 5.3.2.2 at the Advanced Light Source and the Soft X-ray Spectromicroscopy beamline at the Canadian Light Source, as detailed in the SI. The lower Fe and Cr concentrations of the montmorillonite reaction products did not allow them to be imaged with STXM, nor did scanning electron microscopy allow the distribution or morphology of Cr products to be characterized in detail due to the low electrical conductivity of both clay minerals.

Chromium(VI) Reduction Kinetics and Pathway
Aqueous Cr(VI) concentrations decreased in every reactor containing reduced clay minerals.
Aqueous Fe(II) concentrations were below the ferrozine detection limit (~0.1 μM), and control reactors containing Cr(VI), buffer, and unreduced clay minerals showed no significant decreases in aqueous Cr(VI) concentrations from reduction by the buffer or sorption by the clay minerals ( Figure S2). The kinetics are consistent with a second-order rate law (eq 1, Figure where k is the second-order rate constant (M -1 min -1 ). The last term represents the concentration of Fe(II) and is calculated from XFe(II ); fFe, the weight fraction of Fe in the clay mineral (g Fe/g clay mineral); dcl ay, the clay mineral suspension density (g clay mineral/L); and MWFe, the molecular weight of Fe (g/mol). Changes in XFe(II) over the course of the reaction were calculated by assuming a stoichiometric 3:1 reaction with the lost Cr(VI).
Due to the excess of Fe(II) necessary to minimize changes in XFe(II) , the kinetics approach a pseudofirst-order model with respect to Cr(VI) ( Figure S3). However, pseudo-first-order rate constants k1 (min -1 ) cannot be used to compare clay minerals with different values of XFe(II) because the concentration of Fe(II), which is embedded in k1, depends on XFe(II) (eq 1). When ρcl ay is halved, k1 also decreases by roughly half, which suggests that the reaction is first order with respect to Fe(II) and second order overall ( Figure S4). Thus, to allow comparison amongst different XFe(II) values, second-order rate constants are used in the remainder of the text.
Second-order kinetics are consistent with previous studies of Fe(II/III)-bearing clay minerals. A second-order rate law was observed for Cr(VI) reduction by biotically reduced clay minerals 16 and nitroaromatic reduction by a low-Fe, abiotically reduced montmorillonite. 63 Nitroaromatic and organochlorine reduction by a Fe-rich, abiotically reduced smectite followed a more complex, biphasic second-order kinetic model involving two interconvertible Fe(II) species of differing reactivity. 63,64 Although the abiotically reduced nontronite may contain multiple types of Fe(II) sites with different reactivities towards Cr(VI), 38,45 multi-phase kinetics were not observed in our experiments: R 2 values for a single-species, second-order model are greater than 0.9 under all conditions (Tables S1 and S2).
For highly reduced clay minerals, more Cr(VI) is lost in the initial seconds before the first sampling point than predicted by the second-order kinetics that describe the rest of the reaction.
Although previously attributed to sorption, 22 this lost Cr(VI) is more likely reduced by residual Fe(II)-containing complexes sorbed on the clay surface, which is one of the fastest reductants of Cr(VI). 31 Surface-complexed Fe(II) may derive from the reduction of the clay minerals by sodium dithionite, which causes minor release of Fe(II) from the clay mineral structure. 57 Because sorbed Fe(II) reacts with Cr(VI) much faster than Fe(II) in the clay mineral structure, it is swiftly depleted and does not appear to compete with structural Fe(II) over the course of the reaction beyond the initial sampling point. The dominant reductant beyond the first sampling point in all reactors is therefore assumed to be structural Fe(II). All structural Fe in the montmorillonite is octahedral. 42 As much as 8% of Fe in the nontronite structure is tetrahedral. 45 This tetrahedral Fe may be redoxactive 33 but is likely not the major reductant of Cr(VI) for the nontronite, as discussed below.
Under our experimental conditions, reduction of all Cr(VI) by Fe(II) at edge sites is impossible without intra-layer electron hopping. If all Cr(VI) is edge-sorbed and no intra-layer electron hopping occurs, a concentric reaction front moving towards the interior of the clay mineral particle should develop. Dynamic light scattering indicates that clay mineral particles are somewhat larger than their nominal maximum of 2 μm under experimental conditions, likely due to aggregation. A conservative back-of-the-envelope calculation assuming reaction of 5% of a flat circular particle 2 μm in diameter implies that the distance of electron transfer would increase by the end of the reaction to more than 250 Å, compared to typical outer-sphere electron transfer distances of less than 15 Å in aqueous solutions. 65 Electron hopping over such large distances between the isolated Fe sites in the montmorillonite is expected to be extremely slow. However, extensive electron transfer remains possible for the nontronite because electron hopping between adjacent Fe(II/III) sites in the octahedral layer 33,51 is fast (~10 5 transfers/s). 32 Under these conditions, Cr(VI) therefore must primarily sorb on the montmorillonite basal plane to be reduced, but Cr(VI) sorbed on the nontronite edges or basal planes may be reduced.
To further constrain the reaction pathway, ionic strength and pH were varied. For the montmorillonite, inner-sphere Cr(VI) sorption on the basal plane is unlikely due to the negative charge on CrO4 2-or HCrO4 -and high oxidation state of Cr(VI). An O(-II) bridge between tetrahedral Cr(VI) and tetrahedral Si(IV) or Al(III) would carry an excess of positive charge in violation of Pauling's second rule, although an inner-sphere bond between Cr(VI) and tetrahedral Fe(II) is theoretically possible. 66 Consistent with outer-sphere sorption of Cr(VI), increasing the ionic strength slows the reaction (Figure 2A). No significant differences were seen amongst the background electrolytes, which included SO4 2-and Clanions and Na + and Ca 2+ cations. The indifference to the anion identity suggests that the decrease in k is not due to competition for innersphere sorption sites, which are in any case unlikely to be favorable for sorption. 67 The indifference to the cation identity (Na + vs. Ca 2+ ) suggests that the decrease in k is not caused by collapse of the electrical double layer, aggregation of the clay mineral particles, or decreased clay mineral dissolution and release of aqueous Fe 2+ . Ca 2+ causes greater changes to the double layer, more clay mineral aggregation, and less dissolution than Na + . 67 Furthermore, if the reaction is controlled by surface area, then lowering the pH, which is also expected to increase aggregation and decrease dissolution, should in turn decrease the reaction rate.
Instead, pH has no consistent effect on k, even though HCrO4 -has a higher E° (1.2 V) than CrO4 2-(-0.12 V) and the edge charge changes from negative at pH 7.3 to positive at pH 5.5 ( Figure   2B). 56,68 The lack of a pH effect is consistent with outer-sphere basal plane sorption, in contrast with previous studies that found exclusive Cr(VI) sorption to biotite and phlogopite edge sites. 27,28,18 Despite the electrostatic unfavorability of anionic Cr(VI) sorption to the negatively charged basal plane, it remains possible if the chemical potential of sorption is great. 69 Slight outersphere sorption of anionic Cr(VI) to negatively charged surfaces has been observed, 70 persist long enough on the clay surface for Cr disproportionation. As discussed in the next section, the collective results imply that the rate-determining step follows the first electron transfer.
For the nontronite, increasing the ionic strength slows the reaction to an even greater extent and does not depend on the identity of the background electrolyte (Figure 3A). Similarly to the montmorillonite, the decrease of k is thus consistent with outer-sphere sorption. 67 Cr(VI) reduction is faster at pH 5.5 than at pH 7.3 at every XFe(II) tested ( Figure 3B). Reduction may be faster at the lower pH because HCrO4 -has a higher E° than CrO4 2-, 68 but this does not affect k for the montmorillonite, and E°ef f of the nontronite is also expected to increase with decreasing pH. 72 A more likely explanation is that a significant fraction of the anionic Cr(VI) sorbs as an outer-sphere complex on the clay mineral edge, which is negatively charged at pH 7.3 but positively charged at pH 5.5. 57 Sorption of anionic Cr(VI) is hence more electrostatically favorable at the lower pH.
Cr(VI) sorption on the basal plane followed by reduction by either octahedral or tetrahedral Fe(II) cannot be ruled out but is unlikely to lead to the observed pH dependence, since the diffuse negative charge on the basal plane does not depend on pH. Tetrahedral Fe is far less abundant than octahedral Fe and so is unlikely to be the major reductant of outer-sphere edge-sorbed Cr(VI).

Linear Free Energy Relationships
The dominant influence on reaction kinetics for both clay minerals is the average XFe(II) as expressed by E°ef f. Ionic strength and pH change k by less than an order of magnitude, whereas altering XFe(II) causes k to vary by three orders of magnitude (Tables S1 and S2). The dependence of k on E°ef f can be described with a linear free energy relationship between log(k) and E°ef f for both clay minerals at pH 7.3 and 5.5, although the correlation is somewhat stronger for the montmorillonite (Figure 4) For both clay minerals, the observed slope is smaller than the ideal value for a single electron transfer of -16.9 V -1 : -9.4 ± 0.7 V -1 at pH 7.3 and -7.5 ± 1 V -1 at pH 5.5 for the montmorillonite and or Cr(IV) after electron transfer, or structural alterations of edge-site Fe(II/III) that do not involve electron transfer. 55,77 Although the precise mechanism is unclear, the linear free energy relationship allows the kinetics of Cr(VI) reduction to be modeled over a spectrum of E 0 eff values.

Products
Solid products were investigated using Cr and Fe K-edge XANES spectroscopy for both clay minerals under all pH and ionic strength conditions and a variety of buffer concentrations and Cr(aq):Fe(clay mineral) ratios. All Cr K-edge XANES spectra we measured are consistent with that of a Cr(III)-hydroxide, possibly admixed with Fe(III) (Figures 5, S5, S6). The absent preedge feature at 5993 eV and the strong absorption maximum at 6009 eV show that no significant

Cr(VI) is present in the solids and thus that sorbed Cr(VI) is reduced to Cr(III). Neither XFe(II) nor
pH affects the products (Figures 5 and S5). Lowering the initial Cr(aq):Fe(clay mineral) ratio accentuates the dip in the main edge at 6014.5 eV, which may qualitatively indicate that the Cr(III)hydroxide is enriched in Fe(III) (Figure S6). 22 Potential sources of Fe(III) in the Cr(III)-hydroxide include surface-sorbed Fe(II) and minor clay mineral dissolution during reaction.
Iron K-edge XANES spectra demonstrate that Fe(III) largely remains within the clay mineral structure after reaction (Figure 6). Spectra of the reduced clay minerals after reaction with Cr(VI) can be fit with linear combinations of the spectra for the unaltered, fully oxidized clay mineral and the reduced clay mineral prior to reaction with Cr(VI). The fraction of Fe in the products that is re-oxidized is roughly stoichiometrically equivalent to the amount of Cr(VI) lost from solution (Table S3). Redox cycling of Fe-rich clay minerals causes internal rearrangements of Fe among octahedral sites and (de)protonation reactions. 51,77 In addition, high levels of abiotic reduction can release as much as 35% of total Fe from the clay mineral structure. 78 To assess whether reoxidation by Cr(VI) similarly releases Fe, the fully reduced nontronite was reacted for two weeks with stoichiometric Cr(VI) at pH 7.3 and 5.5. No Fe(II) or Fe(III) was detected in solution, and despite the near total oxidation of Fe(II) in the clay minerals, neither Fe(III)-oxyhydroxides nor any other Fe-bearing phases were detected. Iron K-edge XANES spectra of these reacted solids can be fit with linear combinations of the spectra for the oxidized and reduced nontronite endmembers, which imply that 100% of Fe(II) the nontronite was oxidized at pH 7.3 and 73% at pH 5.5 ( Figure S7). The high surface area of clay minerals likely makes Fe-bearing clay minerals less susceptible to surface passivation by the Cr(III)-hydroxide product than other solid reductants such as magnetite. 13,15

Implications for Modeling of Reaction Kinetics
The rate of Cr(VI) reduction by Fe(II/III)-bearing clay minerals is primarily controlled by XFe(II), which may simplify geochemical modeling of clay mineral-Cr(VI) systems. The Fe content of the clay mineral, pH, and ionic strength have only minor effects on the rate constant k, even though the high Fe content of the nontronite allows Cr(VI) reduction to follow a different reaction pathway. Cr(VI) reduction by the montmorillonite likely proceeds at basal surface sites as outersphere complexes, whereas Cr(VI) reduction by the nontronite likely proceeds at edge sites as outer-sphere complexes (Figure 7). Nevertheless, despite the different reaction mechanisms, ratedetermining steps, and relationships between XFe(II) and E°ef f, the relationship between XFe(II) and k is almost identical for both clay minerals ( Figure S8).
To illustrate the effects of XFe(II) , the kinetics of Cr(VI) reduction were modeled for a well-mixed system at pH 5.5 that contains two reductants, aqueous Fe(II) and the reduced montmorillonite

ASSOCIATED CONTENT
The following file is available free of charge.
Details about control reactors, kinetic modeling, and XANES and STXM methods; tabulated rate constants; additional Cr and Fe K-edge XANES spectra; STXM images (pdf)  Tables  Table S1. Second-order rate constants for Cr(VI) reduction by nontronite. Table S2. Second-order rate constants for Cr(VI) reduction by montmorillonite. Table S3. Reaction stoichiometry. Figure S1. Fe XANES spectra of reduced clay minerals. Figure S2. Aqueous Cr(VI) concentrations in control reactors containing oxidized clay minerals. Figure S3. Aqueous Cr(VI) concentrations and pseudo-first-order kinetics. Figure S4. Comparison of pseudo-first-order kinetics. Figure S5. Cr XANES spectra showing the effects of pH on the Cr products. Figure S6. Cr XANES spectra showing the effects of Cr(aq):Fe(clay mineral) ratio on the Cr products. Figure S7. Fe XANES spectra for the fully reduced nontronite after 2 weeks of reaction with stoichiometric Cr(VI). Figure S8. Second order rate constants of Cr(VI) reduction as a function of average X Fe(II) . Figure S9. STXM images of abiotically reduced nontronite reacted with Cr(VI). Figure S10. STXM image of the abiotically reduced nontronite.

Control Reactors
Control reactors containing 10 μM Cr(VI) and 0.5 mM HEPES or acetate showed no decrease in Cr(VI) concentrations over the course of a week. Similarly, reactors containing 10-100 μM Cr(VI) and oxidized clays did not show significant decreases in Cr(VI) concentrations over the course of a week (Figure S2).

Kinetic Modeling
A second-order rate law was assumed in order to compare clays with different values of X Fe (II) . A derivation of the linearized rate law is presented below: (1) where the last term will be denoted as [Fe(II)] for convenience. The rate law corresponds to the nominal elementary reaction Cr(VI) + Fe(II) → intermediates + products and can be solved using the dummy variable x for the progress of the elementary reaction: The left side can be integrated using the method of partial fractions: where E ø h and β are constants tabulated in . 2 Combining these equations,

X-Ray Absorption Near Edge Structure Spectroscopy
Chromium and Fe XANES spectroscopy was performed at beamlines 4-1 and 11-2 at the Stanford Synchrotron Radiation Lightsource using a Si (220) crystal at ϕ = 90°. After reaction, samples were centrifuged, and the partially dried solids were sandwiched between Cr-and Fefree Kapton tape and mounted at room temperature in a He-purged sample chamber. The beam was detuned by 50% to minimize harmonics. A Ge array detector was used to measure sample fluorescence, and XANES spectra were collected in 0.2 eV increments around the Cr and Fe absorption edges. Neither a filter nor Soller slits was used for Cr spectra due to the low signal. The K-edge spectrum of Cr metal foil was taken simultaneously with every sample, and the first maximum of the first derivative was calibrated to 5989.0 eV. For Fe, a Mn filter and Soller slits were used to reduce signal noise. The K-edge spectrum of Fe metal foil was taken simultaneously with every sample, and the first maximum of the first derivative was calibrated to 7112.0 eV. No changes in Fe or Cr spectra after repeated scanning were observed, indicating that beam damage was minimal.
Scanning Transmission X-Ray Microscopy STXM was performed at beamline 5.3.2.2 at the Advanced Light Source and the Soft X-ray Spectromicroscopy beamline at the Canadian Light Source. Iron and Cr maps were collected below the L2 edge for each element (700.0 and 570.0 eV, respectively) and at the L2 maximumintensity energy for each oxidation state (708.0 and 710.0 eV for Fe(II) and Fe(III), respectively; 578.0 and 580.5 eV for Cr(III) and Cr(VI), respectively). 3,4 Maps were iteratively aligned and converted from transmission to optical density. For both Fe and Cr, the below-edge background map was subtracted from each above-edge map, generating maps of the relative intensities for each oxidation state.
STXM images of the partially reacted nontronite reveal that Cr(III) is evenly distributed on the basal surface of the clay particles ( Figure S9). The particle edges do not show any buildup of Cr(III), whose migration to the basal plane may be assisted by complexation with the HEPES or acetate buffers. Precipitation on the basal surface or in the interlayer space instead of at the edges is consistent with the lack of passivation of the nontronite particle surfaces, which would be expected if Cr(III) precipitated at the reaction sites. 5,6 The distribution of Fe(II/III) in the reacted nontronite is more ambiguous: Fe(III) appears to be concentrated in areas of greater clay density, but this may be an artifact of image processing. Bulk Fe XANES spectra do not support the formation of Fe(III)-rich phases. In the reduced nontronite not yet reacted with Cr(VI), Fe(II) and Fe(III) are evenly distributed (Figure S10).     were buffered by sodium HEPES, and samples at pH 5.5 were buffered by sodium acetate. Cr(aq):Fe(clay mineral) is 0.033 for both montmorillonite spectra and 0.017 for both nontronite spectra. Both clays were fully reduced prior to reaction with Cr(VI). Figure S6. Cr XANES spectra showing the effects of Cr(aq):Fe(clay mineral) on the Cr products. All reactors were buffered at pH 7.3. Figure S7. Fe XANES spectra for the fully reduced nontronite after 2 weeks of reaction with stoichiometric Cr(VI). The sample at pH 7.3 was buffered by sodium HEPES, and the sample at pH 5.5 was buffered by sodium acetate. Dotted black lines show linear combination fits. Figure S8. Second-order rate constants of Cr(VI) reduction as a function of average X Fe (II) Figure S10. STXM image of the abiotically reduced nontronite showing A) optical density and B) distribution of Fe(II) (blue) and Fe(III) (red) for the same area. Blue-red scale is arbitrary.