Oxidative alteration of ferrous smectites and implications for the redox evolution of early Mars

Abstract

Surface conditions on early Mars were likely anoxic, similar to early Earth, but the timing of the evolution to oxic conditions characteristic of contemporary Mars is unresolved. Ferrous trioctahedral smectites are the thermodynamically predicted products of anoxic basalt weathering, but orbital analyses of Noachian-aged terrains find primarily Fe³⁺-bearing clay minerals. Rover-based detection of Fe²⁺-bearing trioctahedral smectites at Gale Crater suggest that ferrous smectites are the unoxidized progenitors of orbitally-detected ferric smectites. To assess this pathway, we conducted ambient-temperature oxidative alteration experiments on four synthetic ferrous smectites having molar Fe/(Mg+Fe) from 1.00 to 0.33. Smectite suspension in air-saturated solutions produced incomplete oxidation (24–38% Fe³⁺/ΣFe). Additional smectite oxidation occurred upon re-exposure to air-saturated solutions after anoxic hydrothermal recrystallization, which accelerated cation and charge redistribution in the octahedral sheet. Oxidation was accompanied by contraction of the octahedral sheet (d₍₀₆₀₎ decreased from 1.53–1.56 Å to 1.52 Å), consistent with a shift towards dioctahedral structure. Ferrous smectite oxidation by aqueous hydrogen peroxide solutions resulted in nearly complete Fe²⁺ oxidation but also led to partial Fe³⁺ ejection from the structure, producing nanoparticulate hematite. Reflectance spectra of oxidized smectites were characterized by (Fe³⁺,Mg)₂-OH bands at 2.28–2.30 μm, consistent with oxidative formation of dioctahedral nontronite. Accordingly, ferrous smectites are plausible precursors to observed ferric smectites on Mars, and their presence in late-Noachian sedimentary units suggests that anoxic conditions may have persisted on Mars beyond the Noachian.

1. Introduction

The present Martian atmosphere is oxic, but the temporal evolution of its redox state from primordial anoxic conditions is unresolved. Some planetary evolution models (Chevrier et al., 2007) suggest that photochemical water splitting and hydrogen escape were prevalent and produced an oxidizing atmosphere as early as the Noachian (approximately 4.1 to 3.7 Gyr ago (Carr & Head, 2010)). In contrast, climate models accounting for heavy volcanic outgassing in the Noachian predict an anoxic Martian atmosphere (Catling & Moore, 2003; Ramirez et al., 2014; Sholes et al., 2017) where volcanogenic gases (H₂S, SO₂, H₂, CO, and CO₂) would have efficiently consumed photochemically produced oxidants. Such anoxic conditions could theoretically have persisted as long as vigorous volcanic activity continued, i.e. into the Hesperian (approximately 3.7 to 3.1 Gyr ago) (Sholes et al., 2017). Petrologic studies of the Noachian-aged Martian meteorite ALH84001 (King & McSween, 2005; Righter et al., 2008) also require anoxic or minimally oxic conditions in the surface and near-subsurface.

The extensive record of H₂O/OH-bearing minerals at the Martian surface, observed in situ by rover-based methods and from orbit using visible/near-infrared (VNIR) reflectance spectroscopy, provides constraints on the timing of Martian oxidation. Fe/Mg-bearing phyllosilicates, proposed to form by fluid interaction with basaltic crust in surface and subsurface environments, constitute 89% of exposures of H₂O/OH-bearing phases (Carter et al., 2013; Ehlmann & Edwards, 2014). The thermodynamically favored products of basalt weathering under anoxic conditions are Fe²⁺- and Mg-bearing smectites (Burns, 1993; Catalano, 2013). Fe²⁺-Mg smectite formation by aqueous alteration of ferromagnesian minerals has been demonstrated experimentally to occur only under anoxic conditions (Dehouck et al., 2014; Dehouck et al., 2016; Nickerson et al., 2016). Therefore, the detection of a trioctahedral, Fe²⁺-bearing smectite in sediments analyzed by the Curiosity rover at Yellowknife Bay in Gale Crater (Vaniman et al., 2014; Treiman et al., 2014; Chemtob et al., 2015) potentially supports anoxic conditions at the time of their formation in the late Noachian/early Hesperian.

In contrast to the Fe²⁺-bearing smectites observed in situ at Gale Crater, most orbitally detected Fe-bearing phyllosilicates, including those at Gale Crater, are interpreted to be ferric dioctahedral smectites, e.g., nontronite (Mustard et al., 2008; Poulet et al., 2008; Murchie et al., 2009; Milliken et al., 2010; Poulet et al., 2014). Some nontronite detections, such as those at Valles Marineris, are associated with thick sedimentary units, suggesting subsurface phyllosilicate formation (Roach et al., 2010; Ehlmann et al., 2011). Dioctahedral smectites have also recently been detected in situ by Curiosity at Gale Crater in the Murray Formation, upsection from Yellowknife Bay (Bristow et al., 2017).

One available explanation for the prevalence of Martian nontronite is that the smectite is the initial (direct) product of aqueous alteration of Fe²⁺-rich basalt. This nontronite formation pathway implies co-located chemical oxidants and basalt at the time of oxic alteration. Alternatively, ferrous iron in basalt can oxidize under anoxic conditions during water-rock interaction by reduction of OH/H₂O to H₂, as occurs during serpentinization (Oze & Sharma, 2007). However, formation of fully ferric smectite in anoxic environments is rare; dominant minerals produced during serpentinization are magnetite, Fe³⁺-bearing serpentine, and, at low temperatures, ferrous or ferroan smectites (Evans et al., 2013; Dehouck et al., 2014; Bristow et al., 2015).

We consider here an alternative scenario in which Fe³⁺-bearing smectites observed from orbit and detected by Curiosity are not initial alteration phases but are instead the products of oxidation of precursor ferrous smectites formed by anoxic aqueous alteration (Catalano, 2013). Ferrous smectites occur on Earth in altered oceanic crust below the zone in which oxygenated seawater penetrates (Andrews, 1980; Badaut et al., 1985; Alt et al., 1986; Alt & Teagle, 2003) and as alteration products of continental volcanic units (Koyama et al., 1973; Treiman et al., 2014), but are difficult to study because of their instability in the terrestrial atmosphere. Consequently, the spectroscopic properties of ferrous smectites and their oxidation products have been the subject of few studies (Chukanov et al., 2003; Treiman et al., 2014; Chemtob et al., 2015). In the modern Martian near-surface environment, oxidants that could drive ferrous smectite oxidation include O₂, ozone (O₃) and hydrogen peroxide (H₂O₂), which are present at low concentrations in the atmosphere (Encrenaz et al., 2012; Mahaffy et al., 2013) and other photochemically produced oxidants, including nitrate (NO₃⁻) and perchlorate (ClO₄⁻), that occur in detectable concentrations in the Martian soil (Klein et al., 1976; Zent & McKay, 1994; Hecht et al., 2009; Kounaves et al., 2010). While oxidative alteration of ferrous smectites to Fe³⁺-rich smectites like nontronite is thermodynamically favorable for such species (Catalano, 2013), it has not been established experimentally that these Mars-relevant oxidants will generate ferric smectites from ferrous precursors.

The mechanisms involved in the oxidation of trioctahedral ferrous smectites are poorly characterized in the literature, although structural modifications that occur during reduction of nontronite have been extensively studied (e.g., Stucki & Roth, 1977; Komadel et al., 1995; Manceau et al., 2000; Neumann et al., 2011). Briefly, reduction of Fe³⁺ in ferric smectites to Fe²⁺ is accompanied by Fe migration from cis-octahedral sites to adjacent trans-octahedral sites, forming local trioctahedral domains (Manceau et al., 2000; Neumann et al., 2011). Fe reduction and migration is often accompanied by dehydroxylation, and structural rearrangements are only partially reversible upon reoxidation, depending on the Fe content of the smectite and the extent of dehydroxylation (Komadel et al., 1995; Neumann et al., 2011). However, it is unclear from previous studies whether these mechanisms also operate during the oxidation of initially trioctahedral ferrous smectites, which contain few octahedral vacancies to accommodate Fe migration. Instead, Fe ejection from the octahedral sheet may be required to facilitate oxidation (Badaut et al., 1985; Decarreau & Bonnin, 1986). In addition, smectite layer charge behavior is not analogous between reduced nontronites and native trioctahedral ferrous smectites. Reduction of Fe³⁺ in nontronites will increase the layer charge, producing a high-charge phyllosilicate. In contrast, oxidation of Fe²⁺ in trioctahedral smectites will lower and possibly neutralize the layer charge. Structural differences between dioctahedral nontronites and trioctahedral ferrous smectites likely produce substantially different redox behaviors.

In this study, ferrous smectites were synthesized and aqueously altered in the presence of two oxidants relevant to current Martian surface conditions, namely O₂ and H₂O₂. Although detected on Mars, nitrate was not chosen as an oxidant for study because NO₃⁻ has been shown to not oxidize Fe²⁺ in reduced nontronite (Zhao et al., 2015) or ferrous smectite (Beehr & Catalano, 2012). Similarly, perchlorate was not studied because ClO₄⁻ is not known to react abiotically with dissolved or solid phase Fe²⁺ (Urbansky, 1998). Here we characterize the effects of O₂- and H₂O₂-bearing aqueous solution at terrestrial ambient temperatures on smectite structure, Fe coordination and redox state, and VNIR reflectance spectral signatures.

2. Methods

2.1. Materials and experimental setup

Four ferrous smectite clay minerals (labeled smectites A, B, C, and D) with Fe/(Mg+Fe) molar ratios from 1.00 to 0.33 were studied (Table 1). The smectites were synthesized using a sol-gel method (Decarreau & Bonnin, 1986) that was modified to prevent oxidation of iron (Chemtob et al., 2015). Briefly, amorphous silicate gels were precipitated from reagent solutions in a Coy vinyl anaerobic chamber (97% N₂, 3% H₂, p(O₂) < 1 ppm). The anaerobic atmosphere is maintained using palladium catalyst packs. After precipitation, the gels were washed to remove excess salts and then hydrothermally aged at 200° C for 15 days in a vacuum oven, backfilled with N₂ to maintain anaerobic conditions. X-ray diffraction (XRD) and X-ray absorption near-edge structure (XANES) spectroscopy confirmed that the products are smectites with Fe³⁺/ΣFe ≤ 0.06, basal (001) spacing between 14 and 16 Å, turbostratic stacking, (060) spacing between 1.53 and 1.56 Å, and unit cell size increasing with increasing octahedral Fe content. The values for (060) d-spacing correspond to trioctahedral stoichiometry (Moore & Reynolds, 1989). Inductively-coupled plasma optical emission spectrometry (ICP-OES) analyses of digested smectites indicated octahedral site occupancies between 2.8 and 3.0.

Table 1.

Smectite compositions and Fe redox state, as determined by XANES and Mössbauer spectroscopy.

Smectitea	Composition	Treatment	Fe3+/ΣFeb
			XANESc	Mössbauere	Minimum
A (1.00)	Ca0.17(Fe2+2.58Fe3+0.08Al0.27)VI(Si3.54Al0.46)IVO10(OH)2	unaltered	0.03	0.00	0.00
		air-oxidized	0.57	0.24	0.24
		re-crystallized	0.41	0.38	0.38
		re-oxidized	0.94	0.94	0.94
		peroxide	n.d.d	1.00	1.00
B (0.90)	Ca0.18(Fe2+2.36Mg0.27Al0.28)VI(Si3.54Al0.46)IVO10(OH)2	unaltered	0.00c	0.00	0.00
		air-oxidized	0.38	0.80	0.38
		peroxide	n.d.d	1.00	1.00
C (0.63)	Ca0.21(Fe2+1.63Fe3+0.09Mg1.01Al0.22)VI(Si3.46Al0.54)IVO10(OH)2	unaltered	0.05	0.16	0.05
		air-oxidized	0.27	0.30	0.27
		re-crystallized	0.45	0.32	0.32
		re-oxidized	0.43	0.42	0.42
		peroxide	1.00	1.00	1.00
D (0.33)	Ca0.18(Fe2+0.87Mg1.75Al0.28)VI(Si3.53Al0.47)IVO10(OH)2	unaltered	0.00	0.05	0.00
		air-oxidized	0.24	0.40	0.24
		peroxide	1.00	0.97	0.97

^aMolar ratio Fe/(Mg+Fe) is given in parentheses.

^bUncertainty in Fe³⁺/ΣFe is ±0.01 for XANES and ±0.02 for Mössbauer.

^cXANES spectra of unaltered smectite B and nontronite NAu-2 were used as standards for linear combination fitting (LCF); their Fe³⁺/ΣFe values were defined as 0.00 and 1.00, respectively. Fe³⁺/ΣFe for other samples was determined by LCF.

^dNo XAFS spectra were collected for peroxide-treated A or B.

^eValues for unaltered samples from Chemtob et al. (2015)

In air oxidation experiments, 200 to 500 mg of powdered smectite was suspended in ~100 mL of deoxygenated 0.5 m CaCl₂ solution in the anaerobic chamber. The suspensions were removed from the glovebox and placed on the benchtop under ambient lab conditions (T = 22° C). Capillary tubing connected to a small pump was fed to the bottom of the bottle. Water vapor saturated laboratory air (P_O2 = 0.21) was vigorously bubbled through the solution for 7 days. The CaCl₂ solution was used to encourage continuous occupation of the smectite interlayer site by Ca during oxidation and to promote stacking. At the end of the oxidation period, the smectite was returned to the anaerobic chamber, transferred to centrifuge tubes, and separated from the oxygenated solution by centrifugation. After several cycles of washing in deionized, deoxygenated water, the reacted smectites were allowed to air-dry in the nominally anaerobic chamber.

Fe²⁺-bearing minerals subjected to oxidation then subsequently removed from exposure to the oxidant may continue to undergo transformations due to structural relaxation; however, the timescale of those transformations at ambient temperatures may exceed that which can be feasibly observed experimentally. To accelerate relaxation of disrupted structures post-oxidation, subsamples of two air-oxidized smectites (A and C) were hydrothermally treated at 200° C (recrystallized) for 7 days under anoxic conditions. Recrystallized smectites were then further oxidized in air-saturated solution (reoxidized) for an additional 7 days.

In hydrogen peroxide oxidation experiments, 120 to 160 mg of powdered smectite was introduced to ~25 mL of a 10 mM CaCl₂ solution with 0.2% H₂O₂ for 3 days. After treatment, the reacted smectites were separated from the peroxide solution by centrifugation. Preliminary experiments showed that peroxide oxidation caused considerable structural damage to the smectites (Beehr & Catalano, 2012). To characterize the structures of products post-relaxation, the peroxide-treated smectites were hydrothermally recrystallized in anoxic conditions at 200° C for 7 days.

Post-synthesis oxidation of ferrous smectites in nominally anaerobic chambers is difficult to entirely prevent even at low O₂ concentrations if the reaction is kinetically favored, and we did encounter post-synthesis oxidation for some samples during Mössbauer measurements as previously reported (Chemtob et al., 2015). We also encountered unintentional oxidation under nominally anaerobic conditions in connection with this study (section 3.1); this oxidation complicated but did not obscure interpretation of results related to oxidation by aqueous air-saturated and H₂O₂ solutions.

2.2. Analytical methods

XRD measurements were performed using a Bruker d8 Advance powder diffractometer with a Cu Kα X-ray source and a LYNXEYE XE silicon strip detector. Under anaerobic conditions, smectites were loaded into a sample holder consisting of a recess in a silicon zero diffraction plate, then enclosed within an airtight dome to prevent oxidation. The dome is transparent to the X-ray beam, but displays a broad diffraction feature over the interval 8° to 17° 2θ. Scans were performed over a 2θ range of 3° to 80°. Unit cell parameters were determined by nonlinear least squares fitting of observed diffraction peaks using SixPACK (Webb, 2005), assuming a Voigt peak shape with 50% Gaussian and 50% Lorentzian character.

Structural parameters and average Fe oxidation states of smectites and their oxidized products were determined using Fe K-edge X-ray absorption fine structure (XAFS) spectroscopy, conducted at beamlines 5-BM-D, 12-BM-B, and 13-BM-D at the Advanced Photon Source (APS) at Argonne National Laboratory. All beamlines employed Si(111) double-crystal monochromators and beamline-specific mirror configurations for harmonic rejection. Data were collected simultaneously in transmission and fluorescent yield mode, at incident beam energies ranging from 200 eV below the Fe absorption edge (7112 eV) to 800 eV above the absorption edge (corresponding to k = 15 Å⁻¹). This range spanned the XANES region (−30 eV to +50 eV relative to the Fe edge) and extended X-ray absorption fine structure (EXAFS) region (+50 eV to +800 eV). Incident energy was incremented at 5 eV intervals from −200 eV to −50 eV, 1 eV intervals from −50 eV to −10 eV, and 0.25 eV intervals from −10 eV to +30 eV. Above +30 eV, step sized varied with k² to allow additional counting time at high k. Total acquisition time was ~50 minutes per XAFS spectrum.

Modeling procedures of EXAFS spectra of untreated ferrous smectites were previously described in Chemtob et al. (2015). Raw spectral scans were averaged and normalized using Athena, a graphical user interface for IFEFFIT (Newville, 2001). These normalized EXAFS spectra were k³-weighted and analyzed in SixPACK. Scattering paths for structural model fitting were derived from the biotite structure using FEFF 7.02 (Ankudinov et al., 1988). Biotite was chosen as a model structure because it contains trioctahedral sheets hosting ferrous iron. Cation sites in trioctahedral sheets are typically distorted and consist of two short cation-oxygen bonds and four long cation-oxygen bonds; these two paths are labeled Fe-O1a and Fe-O1b, respectively, in our model. Other scattering paths used in modeling include cations in neighboring octahedral sites (Fe-Fe, Fe-Al_oct, and Fe-Mg) and Si in the tetrahedral sheet (Fe-Si). Multiple scattering was not modeled. For each scattering path, the coordination number (N), interatomic distance (R), and a Debye-Waller factor representing disorder (σ²) were determined using nonlinear least squares fitting. For scattering paths from neighboring octahedral sheet cations (Fe-Fe, Fe-Al, and Fe-Mg), values of N were fixed according to smectite stoichiometry. We assumed single values of R and σ² for all three octahedral scattering paths. Coordination numbers (N) for Fe-O1a, Fe-O1b, and Fe-Si were fixed at 2, 4, and 4, respectively.

Modeling of EXAFS spectra of oxidized smectites followed a similar procedure. As above, scattering paths were derived from the biotite structure. We also modeled changes in octahedral site occupancy with oxidation, as might occur with transformation to a dioctahedral structure. For simplicity, we fit an occupancy variable, allowed to vary between 0 and 1, that linearly modified the coordination number for Fe-Fe, Fe-Al, and Fe-Mg. In other words, for the purpose of the EXAFS model, we fixed the octahedral Fe/Mg and Fe/Al ratios for the partially oxidized smectites to be equal to that of unoxidized precursors. We fit the Fe-O peak for the partially oxidized smectites using just two shells, as in the ferrous smectites. In reality, six or more Fe-O shells may be present: two shells for Fe²⁺ in trioctahedral coordination, two shells for Fe³⁺ in dioctahedral coordination, and two shells for Fe³⁺ in ferrihydrite and/or nanohematite. EXAFS cannot resolve these shells; however, this fitting procedure does allow detection of average changes in Fe-O interatomic distance. Similarly, derived Fe-Fe distances represent an average of numerous coordination environments (dioctahedral Fe, trioctahedral Fe, and Fe in iron oxyhydroxides) and can be used to track bulk average changes in Fe coordination.

The XANES spectral region was used for determination of average Fe oxidation state and speciation via linear combination fitting (LCF). XAFS spectra of additional materials were collected for use as endmember standards for LCF of the spectra of oxidized smectites. Smectite standards included nontronite NAu-2 (Fe³⁺/ΣFe = 1) and a synthetic ferrous smectite confirmed by MB to have Fe³⁺/ΣFe = 0 (Chemtob et al., 2015). Because several oxidized smectite samples are shown by XRD to contain Fe oxides or oxyhydroxides (section 3.2), we also synthesized and collected XAFS spectra for two-line ferrihydrite and hematite for use as LCF standards. Ferrihydrite and hematite were each synthesized following methods from Schwertmann & Cornell (2008).

The XANES measurements of Fe oxidation state were complemented by Mössbauer (MB) spectroscopy. MB spectra were collected using a MIMOS-II instrument, located at NASA Johnson Space Center (JSC) and operating in backscatter measurement geometry at 293 K under desiccating conditions (N₂ gas purge with <100 ppm_v H₂O and <100 ppm_v O₂), except for the peroxide-treated samples which were analyzed under ambient laboratory conditions because XANES measurements indicated the samples were fully oxidized. MB analysis postdated XANES analysis by 3 to 24 months (Table S1). Spectra were velocity-calibrated and fit using a least squares fitting procedure described elsewhere (Agresti et al., 2006; Agresti & Gerakines, 2009). Briefly, peak areas and peak widths of doublet pairs were constrained to be equal during fittings. For the spectra that featured hematite sextets, center shift and quadrupole splitting were constrained to literature values (0.37 mm/s and −0.20 mm/s, respectively) (Morris et al., 1995) because of low peak intensities. Where the fitting model included two ferric doublets, the peak width for the two doublets were constrained equal. The recoil-free fraction (f(Fe³⁺)/f(Fe²⁺)) was set to 1.21 after De Grave & Van Alboom (1991) and Morris et al. (1995), who found that value to be valid for a mineralogically diverse range of Fe-bearing samples.

Visible and near-infrared (VNIR) reflectance spectra were collected using Analytical Spectral Devices (ASD) spectrometers at Washington University (WU) and at JSC over the range 0.35 to 2.5 μm. At both sites, spectra were collected using a reflectance probe attachment that contains its own light source; a Spectralon 99% diffuse reflectance standard was also analyzed to correct the smectite spectra to absolute reflectance. WU spectra were collected shortly after the conclusion of each oxidation experiment. VNIR spectra at WU were collected in ambient air (295 K, ~40% RH); subsamples of the smectites were heatsealed during transport from the anaerobic chamber to the spectrometer and were opened to air for a maximum of 3 minutes before analysis. Repeat VNIR analyses over a 2–3 minute period showed no detectable changes in spectral properties over that time. VNIR analyses at JSC were conducted immediately following each MB analysis. At JSC, VNIR spectra were collected under the same desiccating conditions as the MB measurements, except for the peroxide-treated samples, which were analyzed under ambient laboratory conditions (295 K, ~54% relative humidity (RH)).

3. Results

3.1. Characterization of redox state of oxidation products

Peroxide treatment caused nearly-complete Fe²⁺ oxidation (Fe³⁺/ΣFe ≥ 0.97) in all four smectite samples. In the two peroxide-treated smectites for which both XANES and MB data were collected (C and D), redox determinations by the two methods were in excellent agreement (Table 1). Exposure to air-saturated solution resulted in incomplete oxidation of all four ferrous smectites. Modeling of XANES and MB spectra of samples oxidized in air-saturated solution suggests Fe³⁺/ΣFe contents between 0.24 and 0.80 (Tables 1–2; Figures 1–2). However, large discrepancies between XANES-derived and MB-derived Fe³⁺/ΣFe contents were observed for several samples (O₂-oxidized A, O₂-oxidized B, and O₂-oxidized D).

Table 2.

Mössbauer Parameters at Room Temperature.

Sample	Smectite A				Smectite B				Smectite C			Smectite D
Treatment	air-ox	rextll	reox	perox	air-ox	perox/ rextll	air-ox		rextll	reox	perox/ rextll	air-ox	perox/ rextll
Fe3+/ΣFe	0.24(2)	0.38(2)	0.94(2)	1.00(2)	0.80(2)	1.00(2)	0.30(2)		0.32(2)	0.42(2)	1.00(2)	0.40(2)	0.97(2)
Ferric Doublet Fe3D1
CS	0.41(2)	0.36(2)	0.35(2)	0.35(2)	0.37(2)	0.35(2)	0.35(2)		0.41(2)	0.41(2)	0.34(2)	0.40(2)	0.33(2)
QS	0.68(2)	0.72(2)	0.70(2)	0.65(2)	0.79(2)	0.62(2)	0.78(4)		0.57(2)	0.48(4)	0.67(2)	0.66(2)	0.65(2)
FWHM	0.61(2)	0.58(2)	0.54(2)	0.59(2)	0.55(2)	0.58(2)	0.49(2)		0.58(2)	0.59(2)	0.54(2)	0.46(2)	0.51(2)
A	28(2)	31(2)	71(2)	88(3)	56(2)	82(2)	23(2)		36(2)	25(3)	68(3)	27(2)	51(2)
Ferric Doublet Fe3D2
CS	---	---	0.33(2)	0.32(2)	0.34(2)	0.34(2)	0.34(2)		---	0.40(3)	0.35(2)	0.34(2)	0.36(2)
QS	---	---	1.20(2)	1.11(5)	1.35(2)	1.08(3)	1.25(5)		---	0.93(5)	1.13(3)	1.29(2)	1.18(2)
FWHM	---	---	0.54(2)	0.59(2)	0.55(2)	0.58(2)	0.49(2)		---	0.59(2)	0.54(2)	0.46(2)	0.51(2)
A	---	---	24(2)	12(2)	27(2)	18(2)	12(2)		---	22(3)	21(3)	18(2)	34(2)
Ferrous Doublet Fe2D1
CS	1.13(2)	1.18(2)	1.20(2)	---	1.12(2)	---	1.16(2)	1.12(2)		1.12(2)	---	1.11(2)	[1.14]
QS	2.39(3)	2.17(2)	2.16(2)	---	2.05(4)	---	2.21(5)	2.39(2)		2.25(5)	---	2.38(2)	[2.36]
FWHM	0.58(2)	0.65(2)	0.31(2)	---	0.38(2)	---	0.59(2)	0.45(2)		0.50(3)	---	0.42(2)	[0.50]
A	49(5)	32(3)	5(2)	---	5(2)	---	36(8)	42(5)		22(4)	---	27(4)	0(2)
Ferrous Doublet Fe2D2
CS	1.14(2)	1.13(2)	---	---	1.12(2)	---	1.13(2)	1.13(2)		1.13(2)	---	1.15(2)	[1.13]
QS	2.70(2)	2.61(2)	---	---	2.52(2)	---	2.61(2)	2.69(2)		2.63(2)	---	2.66(2)	[2.72]
FWHM	0.34(2)	0.38(2)	---	---	0.38(2)	---	0.36(3)	0.29(2)		0.38(2)	---	0.33(2)	[0.35]
A	24(5)	25(3)	---	---	11(2)	---	30(7)	22(4)		30(4)	---	28(4)	2(2)
Hematite Sextet
Bhf	---	49.1(4)	---	---	---	---	---	---		---	43.6(8)	---	44.8(4)
A	---	12(2)	---	---	---	---	---	---		---	10(3)	---	13(2)

CS = center shift (mm/s) relative to the midpoint of metallic Fe foil at room temperature; QS = quadrupole splitting (mm/s); B_hf = hyperfine field strength (T); Fe³⁺/ΣFe = fraction of total Fe present as Fe³⁺ (corrected for recoil-free fraction, f(Fe³⁺)/f(Fe²⁺) = 1.21); A = fraction subspectral area; FWHM = full width at half-maximum intensity (mm/s); Fixed parameter values (in brackets) for peroxide-treated samples are averages of corresponding entries for samples A through D in Table 5 of Chemtob et al. (2015). Uncertainties on final digit(s) of parameters are indicated in parentheses.

Figure 1.

Fe K-edge XANES spectra of ferrous smectites and their oxidative products, offset for clarity. Fe K-edge positions, defined as the energy at which normalized absorbance is 0.5, are indicated by black dots. Dotted lines at 7120.5 and 7125.3 eV indicate the approximate edge positions corresponding to Fe³⁺/ΣFe = 0 and 1, respectively.

Figure 2.

Backscatter Mössbauer spectra for smectites A, B, C, and D and their oxidation products, offset for clarity. Each spectrum was normalized to a maximum peak intensity of 0.3. Mössbauer spectra of unaltered samples were acquired after some oxidation during storage (cf., Chemtob et al., 2015).

The differences in measured redox state between XANES and MB could be attributable to systematic error in the collection or modeling of one or both spectral datasets, although both techniques have have been used extensively for quantification of Fe oxidation state in silicates (e.g., Bajt et al., 1994; Morris et al., 1995; Wilke et al., 2005). One source of potential error in Fe redox state from XANES linear combination fitting is incorrect selection of spectral endmembers. Fe coordination environment, as well as redox state, affects Fe K-edge position; thus, LCF fitting using standards that do not closely match the Fe speciation of our standards could introduce systematic errors in Fe³⁺/ΣFe. However, our LCF fits show isosbestic behavior relative to selected standards, consistent with negligible mismatch in Fe coordination environment. Incorrect selection of the recoil-free fraction could also cause systematic biases in MB-derived Fe³⁺/ΣFe, but the selected value has previously been shown to be valid for a range of Fe-bearing mineralogical samples (Morris et al., 1995). In any case, differences between the two methods were observed in both directions, and thus do not appear to be the result of any systematic error in measurement.

We instead attribute the differences to inadvertent exposure of some sample aliquots to oxygen during storage and transport. During MB analysis, samples were exposed to the N₂ purge gas which has ~80 ppm O₂. Additionally, the seals on some polyethylene bags used for sample transport from WU to APS and from WU to JSC may have leaked, or O₂ diffusion may have occurred through the walls of the polyethylene bags. Differences in Fe redox state between sample aliquots intended for MB and XANES analysis are corroborated by VNIR reflectance spectroscopy, which we take as evidence that the differences are not tied to MB and XANES data analysis methods. The VNIR spectra collected at WU shortly before XAFS analysis show marked differences in peak position corresponding to differences in oxidation state from the VNIR spectra collected at JSC at the same time as MB (Section 3.3).

Following the interpretation that the smectites were subject to inadvertent oxidation, we interpret both XANES- and MB-derived values of Fe³⁺/ΣFe as upper limits on the extent of Fe²⁺ oxidation at the conclusion of the oxidation experiments. Consequently, we assume the minimum value of Fe³⁺/ΣFe derived from the two methods corresponds to the redox state at time of experimental completion. On this basis, 7 days of exposure of the four smectites to air-saturated solutions resulted in 24–38% Fe²⁺ oxidation. After hydrothermal recrystallization and re-exposure to air-saturated solutions, oxidation to Fe³⁺ was more complete. Re-oxidized smectites A and C had Fe³⁺/ΣFe of 94% and 42%, respectively.

3.2. Structural changes with oxidation

3.2.1. Oxidation of Fe²⁺-rich smectites in air-saturated solution

Oxidation of the most Fe-rich smectites (A and B) in air-saturated solution caused significant lattice contraction, as shown in XRD patterns by the shift of their (060) d-spacings to smaller values, from 1.564 to 1.558 Å and 1.552 to 1.540 Å, respectively (Table 3; Figure 3c). This octahedral sheet contraction is consistent with the oxidation of structural Fe²⁺ to the smaller Fe³⁺ (Manceau et al., 2000). Recrystallization and reoxidation of smectite A led to further contraction of d₍₀₆₀₎ to 1.525 Å, similar to d₍₀₆₀₎ values observed for dioctahedral nontronite (Moore & Reynolds, 1989). EXAFS shell fitting supports the interpretation of a contracting octahedral sheet. Fe²⁺ in unaltered trioctahedral smectites is coordinated in a distorted octahedral site with an average Fe-O distance of ~2.10 Å (Chemtob et al., 2015; Manceau et al., 1998; Gates et al., 2002; Bonnin et al., 1985). For smectites A and B, EXAFS modeling implies that the average Fe-O distance decreased with air oxidation to 2.04 and 2.06 Å, respectively. Distances to neighboring cations in the octahedral and tetrahedral sheet (Fe-(Fe,Mg,Al,Si)) also contracted upon oxidation (Table 4; Figure 4a,b).

Table 3.

XRD d-spacings^a

Smectite	Treatment	d001 (Å)	d02ℓ (Å)b	d060 (Å)
A	unaltered	16.19(8)	4.579(6)	1.5636(2)
	air-oxidized	16.99(5)	4.552(3)	1.5584(2)
	re-crystallized	15.18(4)	4.601(5)	1.5617(2)
	re-oxidized	15.76(3)	4.506(3)	1.5249(2)
	peroxide	15.61(4)	4.519(2)	1.5203(1)
B	unaltered	15.55(2)	4.652(3)	1.5523(2)
	air-oxidized	15.10(1)	4.560(2)	1.5400(1)
	peroxide	15.43(1)	4.513(2)	1.5203(1)
C	unaltered	15.67(1)	4.600(2)	1.5385(1)
	air-oxidized	15.61(1)	4.584(1)	1.5381(1)
	re-crystallized	15.87(1)	4.573(1)	1.5378(1)
	re-oxidized	15.76(1)	4.567(1)	1.5358(1)
	peroxide	14.98(1)	4.530(1)	1.5242(1)
D	unaltered	14.78(1)	4.579(2)	1.5321(1)
	air-oxidized	15.66(1)	4.549(1)	1.5318(1)
	peroxide	15.00(1)	4.522(1)	1.5229(1)

^aUncertainty for final digit is given in parentheses.

^bBecause of the high asymmetry of the (02ℓ) peak, we report as d_(02ℓ) the peak fitting results for the (02ℓ) maximum.

Figure 3.

Cu Kα XRD patterns of smectites and their oxidative products. In (a) and (b), the region from 2θ from 8° to 18° is omitted from the patterns because of a peak produced by the dome of the sample holder. Peaks associated with nanoparticulate hematite are labeled ‘h’. Small sharp peaks originate from minor quartz introduced by grinding sample with an agate mortar and pestle. c) The XRD 060 region for samples A, B, C, and D and their oxidative products, with colors as in a). Dotted lines correspond to the 060 peak position for each unaltered sample. d₀₆₀ (in Å) for unaltered samples are also provided.

Table 4.

Fe K-edge EXAFS structural fitting parameters

Shell			Smectite A				Smectite B		Smectite C					Smectite D
			unaltered	air-ox	re-xtll	re-ox	unaltered	air-ox	unaltered	air-ox	re-xtll	re-ox	perox	unaltered	air-ox	perox
Fe-O1a	Na		2	2	2	2	0	2	2	2	2	2	2	2	2	2
	Rb		2.016(18)f	2.10(13)	1.936(21)	2.084(26)	-	1.963(17)	1.986(19)	1.936(13)	1.951(29)	1.914(13)	2.065(49)	2.011(28)	1.994(30)	2.066(57)
	σ2 c		0.006g	0.015	0.009	0.008	-	0.009	0.004	0.006	0.009	0.006	0.008	0.004	0.006	0.008
Fe-O1b	N		4	4	4	4	6	4	4	4	4	4	4	4	4	4
	Rb		2.140(7)	2.008(38)	2.094(12)	1.943(10)	2.101(7)	2.109(8)	2.124(11)	2.108(10)	2.096(14)	2.080(9)	1.951(11)	2.135(14)	2.132(17)	1.958(12)
	σ2		0.006(1)	0.0149	0.009(1)	0.008(1)	0.009(1)	0.009(1)	0.004(1)	0.006(1)	0.009(2)	0.006(1)	0.008(3)	0.004(2)	0.006(2)	0.008(3)
Fe-Fe	N		5.45	2.6(8)	3.2(1.1)	1.2(7)	4.87	3.6(7)	3.19	3.1(4)	2.6(5)	2.9(4)	1.6(6)	1.80	1.8(2)	0.9(5)
	R		3.177(22)	3.149(17)	3.218(45)	3.018(28)	3.153(14)	3.139(16)	3.115(18)	3.133(17)	3.136(27)	3.112(18)	2.955(51)	3.095(21)	3.105(18)	3.099(92)
	σ2		0.012(2)	0.012i	0.012	0.012	0.012(2)	0.012	0.011(2)	0.011	0.011	0.011	0.011	0.008(2)	0.008	0.008
Fe-Aloct	N		0.55	0.28(9)	0.33(10)	0.1(7)	0.57	0.42(8)	0.38	0.38(5)	0.31(6)	0.35(5)	0.18(7)	0.58	0.6(1)	0.11(6)
	R		3.177h	3.149	3.218	3.018	3.153	3.139	3.115	3.133	3.136	3.112	2.955	3.095	3.105	3.099
	σ2		0.012h	0.012i	0.012	0.012	0.012	0.012	0.011	0.011	0.011	0.011	0.011	0.008	0.008	0.008
Fe-Mg	N		0	0	0	0	0.56	0.41(8)	2.42	2.4(3)	2.0(4)	2.2(3)	1.2(4)	3.63	3.7(4)	2.2(1.3)
	R		-	-	-	-	3.153h	3.139	3.115	3.133	3.136	3.112	2.955	3.095	3.105	3.099
	σ2		-	-	-	-	0.012h	0.012i	0.011	0.011	0.011	0.011	0.011	0.008	0.008	0.008
Fe-Si	N		4	4	4	4	4	4	4	4	4	4	4	4	4	4
	R		3.296(30)	3.278(64)	3.271(35)	3.246(54)	3.304(22)	3.309(29)	3.290(25)	3.276(27)	3.278(21)	3.246(18)	3.252(18)	3.276(27)	3.273(27)	3.226(25)
	σ2		0.011(7)	0.019(7)	0.007	0.021(1)	0.012	0.013(4)	0.010(4)	.011(3)	0.007(2)	0.009(2)	0.008(2)	0.010(4)	0.010(3)	0.008(3)
	ΔE0d		−5.0(1.0)	−3.7 (1.8)	−4.7 (1.5)	−2.9(1.6)	−4.8(0.9)	−4.3(1.7)	−6.1(1.7)	−5.7(1.3)	−3.6(1.7)	−6.1(1.3)	−5.4(1.6)	−6.2(2.3)	−4.2(2.4)	−6.3(2.3)
Fe-Oaverage (Å)e		2.099		2.039	2.041	1.990	2.101	2.060	2.078	2.051	2.048	2.025	1.989	2.094	2.086	1.994

^aCoordination number.

^bInteratomic distance (in Å)

^cDebye-Waller factor (Ų)

^dDifference in edge energy between experiment and theory (eV)

^eAverage Fe-O distance = (N_Fe-O1a *R_Fe-O1a + N_Fe-O1b *R_Fe-O1b) * ⅙

^fStatistical uncertainties on last digit(s) in fit parameters are reported in parentheses (1 standard error). Parameters with no listed uncertainties were not varied in the analyses.

^gThe Debye-Waller factor for the Fe-O1a shell was set equal to that for the Fe-O1b shell.

^hThe R and σ² parameters for Fe-Al and Fe-Mg were set equal to those for the Fe-Fe shell.

ⁱThe Debye-Waller factor for Fe-Fe, Fe-Al, and Fe-Mg in the oxidation products were set equal to those for the Fe-Fe shell in the corresponding unaltered sample.

Figure 4.

Fourier transforms of EXAFS spectra, uncorrected for phase shift (ΔR) and offset for clarity. Modeled fits to the EXAFS spectra (gray series) were calculated using the R range from 1 to 3.4 Å. For each composition, dotted lines corresponding to positions of Fe-O and Fe-(Fe,Mg,Al,Si) shells in untreated smectite are included for reference. k³-weighted EXAFS spectra (prior to Fourier transform) are displayed in Figure S2.

Although a smectite phase is present after all oxidative treatments of the Fe-rich smectites A and B, XRD and EXAFS demonstrate that a significant fraction of Fe³⁺ is ejected during oxidation in air-saturated solution. Peaks associated with nanoparticulate hematite (14 to 20 nm particle size from the Scherrer equation) are apparent in XRD patterns of recrystallized and reoxidized smectite A (Figure 3a). Mössbauer spectra of recrystallized smectite A also indicate the presence of hematite; a sextet is present, consistent with ~12% of total Fe in the sample occurring as Fe³⁺ hematite particles with diameters greater than ~7 nm (e.g. Morris et al., 1995) (Table 2). Nanoparticulate hematite likely formed by recrystallization of an X-ray amorphous Fe³⁺ phase (e.g., ferrihydrite) produced by ejected Fe during oxidation. Fourier transformed EXAFS (FT-EXAFS) data indicates a drop in the magnitude of the Fe-(Fe,Mg,Al,Si) peak with oxidation, consistent with partial Fe ejection (Figure 4; Table 4). After recrystallization and reoxidation, smectite A is characterized by shorter Fe-O distances (~1.99 Å) and a less distorted Fe-O shell consistent with Fe primarily in a dioctahedral sheet.

3.2.2. Oxidation of mixed Fe-Mg smectites in air-saturated solution

Unlike the more Fe²⁺-rich smectites, the mixed Fe-Mg smectites (C and D) showed little change in d₍₀₆₀₎ upon oxidation in air-saturated solution, recrystallization, and reoxidation (Figure 3c; Table 3). The absence of a lattice contraction is likely a consequence of lower initial Fe²⁺ concentrations in the octahedral sheet. That is, changes in Fe ionic radius with oxidation had little effect on crystallographic unit cell size because Fe was a less abundant component. Additionally, reoxidized smectite C had lower Fe³⁺/ΣFe (0.42) than smectite A (0.94), further decreasing the fraction of octahedral cations that underwent bond length contraction.

In contrast to the XRD-derived d₍₀₆₀₎, which was unchanged by air oxidation, the EXAFS-derived average Fe-O distance in smectites C and D decreased (e.g., from 2.08 Å in unoxidized smectite C to 2.05 Å and 2.02 Å, respectively, for oxidized and reoxidized smectite C) (Figure 4; Table 4). These measurements suggest that octahedral Mg controls the overall unit cell dimension but, locally, Fe octahedra contract with oxidation as cation charge increases. The absence of nanoparticulate hematite in the XRD patterns (Figure 3b) and no detectable change in magnitude of the Fe-(Fe,Mg,Al,Si) peak in FT-EXAFS spectra (Figure 4c; Table 4) suggests that minimal Fe³⁺ ejection occurred during air oxidation of the mixed Fe-Mg smectites.

3.2.3. Oxidation of ferrous smectites by hydrogen peroxide solutions

XRD patterns for all four peroxide-treated smectites after recrystallization have values of d₍₀₆₀₎ (1.520 to 1.524 Å) that are consistent with dioctahedral ferric smectites (Table 3). The result implies oxidation by peroxide solutions is more disruptive to the smectite structure than oxidation by air-saturated solutions. After recrystallization, the XRD patterns of all four peroxide-treated smectites contain peaks for nanoparticulate hematite, consistent with Fe³⁺ ejection (Figure 3a–b; Figure S1). Mössbauer spectra of peroxide-treated smectites C and D also display a nanoparticulate hematite sextet, in agreement with XRD patterns (Figure 5; Table 2). EXAFS spectra were collected for peroxide-treated smectites C and D only (Figure S2; Figure 4). Shell fitting for those spectra indicates that, relative to untreated smectites, peroxide-treated smectites have shorter average Fe-O distances (1.99 Å) and a less distorted Fe-O shell consistent with a dioctahedral structure (Figure 4c–d; Table 4). EXAFS shell fits also show a significant loss of neighboring cations with peroxide treatment. In smectite C, the Fe-Fe coordination number decreased with peroxide treatment from 3.2 (for unaltered sample) to 1.6±0.6 (Table 4). This decrease is larger than would be expected solely from a structural transition from trioctahedral to dioctahedral and likely indicates Fe³⁺ ejection from the octahedral sheet.

Figure 5.

Mössbauer spectrum, model fit (red trace), and model subspectra for smectite D that was subjected to hydrogen peroxide treatment. Data and hematite fit are also provided in 2x magnification and offset for clarity.

3.3. Reflectance spectroscopy

VNIR spectral properties of unaltered trioctahedral ferrous smectites were described by Chemtob et al. (2015). Broad absorption bands centered near 0.70, 0.91, and 1.11 μm in the unaltered trioctahedral smectites (Figure 6) result from Fe²⁺ and Fe³⁺ electronic transitions and Fe²⁺-Fe³⁺ charge-transfer transitions. The band near 1.4 μm results from stretching overtones of OH bound to octahedral cations and from the OH stretch in adsorbed and interlayer H₂O. The band near 1.9 μm is produced by stretching plus bending combination vibrations of interlayer and adsorbed H₂O. The features in the 2.1 to 2.5 μm region result from combination and overtone bands of trioctahedral (MMM-OH) and dioctahedral (MM-OH) groups (M = Fe²⁺, Fe³⁺, Mg²⁺, and Al³⁺). The positions of these bands vary with octahedral sheet stoichiometry (Figure 7; Table 5) (Chemtob et al., 2015). Briefly, the band positions assigned to combination vibrations of (Fe²⁺,Mg)₃-OH range from 2.353 μm for the Fe²⁺-rich endmember smectite A to 2.318 μm for the Mg-rich smectite D. The AlAl(Fe²⁺,Mg)-OH feature changes position, from 2.255 μm to 2.245 μm, with decreasing Fe²⁺/Mg ratio. The OH stretching overtone feature also changes position, from 1.418 to 1.405 μm, with decreasing Fe²⁺/Mg ratio.

Figure 6.

VNIR reflectance spectra of smectites A – D and their oxidative products, collected in dry anaerobic conditions, except for the peroxide-treated samples which were measured in laboratory air. Gray dotted lines at ~0.55 μm and 0.87 μm highlight the spectral features of Fe3+ in hematite in peroxide-treated smectites.

Figure 7.

Continuum-removed visible/near-infrared reflectance spectra, offset for clarity. Spectra of O₂-oxidized samples that experienced additional exposure to air during storage and transport are plotted as dotted lines. The spectra of two smectite standards, nontronite NAu-2 and ferric smectite SWa-1, are included for comparison. Vertical lines at the band minima for (Fe³⁺Fe³⁺)-OH and (AlFe³⁺)-OH in nontronite are included for reference.

Table 5.

VNIR OH combination band positions in micrometers^a.

Smectite	Treatment	H2O and M3-OH	AlAl-OH	Al(Fe3+,Mg)-OH	AlAl(Fe2+,Mg)-OH	(Fe3+,Al,Mg)2-OH	(Fe2+,Mg)3-OH	(Fe3+,Mg)2-OHstb	(Fe2+,Mg)3-OHstb
A	unaltered	1.422, 1.377 (sh)	2.195	---	2.255	---	2.353	---	---
	air-oxidized (JSC)c	1.419, 1.376 (sh)	2.197	---	2.258	---	2.356	---	---
	air-oxidized (WU)c	1.419	2.20 (sh)	---	2.259	2.275	---	---	---
	re-crystallized	1.421, 1.375 (sh)	2.197	---	2.261	---	2.369	---	---
	re-oxidized	1.424	2.195 (sh)	2.234	---	2.277	---	2.402	---
	peroxide	1.422	2.19 (sh)	2.240	---	2.284	---	2.408	---
B	unaltered	1.418, 1.376 (sh)	2.195	---	2.252	---	2.346	---	2.428(w)
	air-oxidized (WU)c	1.418	2.194	---	2.250	---	2.337	---	---
	air-ox. (JSC))c	1.418	2.195	---	---	2.272	---	---	---
	peroxide	1.424	---	2.243	---	2.286	---	2.404	---
C	unaltered	1.413, 1.38 (sh, w)	2.191	---	2.249	---	2.325	---	2.413 (w)
	O2-oxidized	1.412	2.195 (sh)	---	2.247	---	2.324	---	2.411
	re-crystallized	1.412	---	---	2.246	---	2.321	---	2.409
	re-oxidized	1.412	---	---	2.247	---	2.317	---	2.408
	peroxide	1.419	---	2.250	---	2.295	---	2.398	---
D	unaltered	1.406	2.196	---	2.245	---	2.318	---	2.400
	air-oxidized (WU)	1.409	2.196 (sh)	---	2.243	---	2.319	---	---
	air-oxidized (JSC)	1.413	2.199	---	---	---	2.331	---	---
	peroxide	1.413	---	---	2.245	2.298	---	2.396	---

^aUncertainty in band positions ±0.003 μm. sh = shoulder, w = weak, and --- = not detected.

^bCombination OH stretch-translation mode.

^cSpectra collected at JSC and WU are both included for those samples in which XANES and MB showed different redox states.

The VNIR spectral properties of ferrous smectites changed significantly upon Fe²⁺ oxidation (Figures 6–7; Table 5). The most oxidized samples (re-oxidized A and all peroxide-treated smectites) are characterized by a band minimum near 0.87 μm, a relative reflectivity maximum near 0.78 μm, an inflection near 0.55 μm, and elevated albedo longward of the ferric absorption edge (~0.55 to ~0.78 μm) (Figure 6; Table 6). These spectral features taken together are characteristic of hematite (e.g., Morris et al., 2000), and we attribute them to the nanoparticulate hematite detected in our XRD and Mössbauer data rather than to smectite Fe³⁺ electronic transitions, which are apparently masked by hematite spectral features. However, the MM-OH and MMM-OH spectral features of these oxidized smectites (discussed below) are not masked because hematite does not have spectral features in that region.

Table 6.

Positions of Fe³⁺ spectral features in peroxide-treated smectites in micrometers^a.

Smectite	Inflection	Relative Reflectivity maximum	Fe3+ absorption minimum
A	0.554	0.770	0.881
B	0.548	0.766	0.875
C	0.547	0.792	0.863
D	0.498b, 0.554	0.788	0.856

^aUncertainty in band positions ±0.005 μm.

^bA second inflection point apparent in peroxide-treated smectite D may be attributable to Fe³⁺ in smectite.

Modest changes in VNIR spectral features near 1.4 and 1.9 μm associated with OH and H₂O were observed with oxidation. For smectites A, B, and C, the OH stretching overtone feature increased in intensity with oxidation in air-saturated solution (Figure S3), but did not change position. A shoulder feature observable at 1.375 μm in smectites A and B, perhaps attributable to a stretching overtone of Si-OH, persisted with oxidation. VNIR spectra of the peroxide-treated smectites were collected in ambient laboratory air, so the hydration features at 1.4 and 1.9 μm are intense and dominated by H₂O; the position of the OH stretching overtone feature varies with Fe³⁺/Mg ratio (Table 5) as discussed next.

Continuum-removed spectra in the 2.1 – 2.5 μm range show that the positions and intensities of the trioctahedral MMM-OH and dioctahedral MM-OH bands also changed in response to oxidation and structural transformation of the smectites (Figure 7). Continuum-removed spectra of the aliquots of air-oxidized Fe-rich smectites A and B that did not experience inadvertent, post synthesis, additional oxidation (i.e. A-air-ox(JSC) and B-air-ox(WU) in Figure 7) are broadly similar to their untreated counterparts, i.e., spectrally dominated by Fe²⁺. The corresponding spectra for smectites A, B, and D that did receive additional oxidation are plotted as dashed lines.

The continuum-removed spectrum of re-oxidized smectite A is distinctly different from its unaltered precursor. The (Fe²⁺)₃-OH band is absent, and an additional band centered near 2.28 μm is observed and assigned to dioctahedral (Fe³⁺)₂-OH consistent with XRD patterns (Table 3) and MB spectra that show Fe is present predominantly as Fe³⁺ (Fe³⁺/ΣFe = 0.94; Table 1). For comparison, corresponding spectra are shown for ferric smectite SWa-1 and nontronite NAu-2 (Figure 7). The spectra of the inadvertently oxidized, Fe³⁺-rich aliquots of air-oxidized smectites A and B (i.e. A-O₂-ox(WU) and B-O₂-ox(JSC)) also lack a (Fe²⁺)₃-OH band and display bands near 2.28 μm that we attribute to (Fe³⁺)₂-OH (dashed lines in Figure 7).

The VNIR spectra of the mixed Fe-Mg smectites C and D after oxidation in air-saturated solution are similar to the spectra of their untreated counterparts. The spectra of recrystallized and reoxidized smectite C also resemble the untreated sample but are characterized by an additional broad shoulder near 2.30 μm (Figure 7). The shoulder is attributed to partial Fe²⁺ oxidation (from Fe³⁺/ΣFe = 0.27 to 0.42; Table 1) with preservation of the trioctahedral structure (Table 3) or a mixed di- and tri-octahedral smectite where the 2.30 μm band is associated with dioctahedral Fe³⁺(Fe³⁺,Mg)-OH (e.g., Bishop et al., 2002). The latter interpretation is supported by the spectrum of peroxide-treated smectite C, which has a well-defined 2.30 μm band (Figure 7; described below).

All untreated and air-oxidized, recrystallized, and re-oxidized smectites have a spectral feature as a shoulder to well-defined minimum near 2.191 μm. A shoulder feature at a similar wavelength was previously observed in spectra of a Fe-rich aluminous smectite by Bishop et al. (2008). This band may represent an OH combination mode associated with Al clusters in the octahedral sheet, and was previously assigned to AlAl-OH by Chemtob et al. (2015). In smectite C, the 2.191 μm feature diminished in intensity with recrystallization and reoxidation.

The spectra of all peroxide-treated smectites feature a strong band located between 2.284 and 2.298 μm that is assigned to dioctahedral Fe³⁺(Fe³⁺,Al,Mg)-OH (Figure 7). The band position increases with increasing Mg/Al ratio (Table 5). The dioctahedral structural assignment also follows from the XRD patterns (Table 3; Figure 3). Therefore, the peroxide-treated smectites are nontronite with variable octahedral contributions from Al and Mg. Although we do not know the relative proportions of Fe³⁺, Mg, and Al because of ejection of Fe³⁺ from the smectite structure during oxidation, the spectra for Mg-poor peroxide-treated A and B are very similar to those for ferric smectite SWa-1 and nontronite NAu-2 (Figure 7). We attribute the shoulder at 2.316 μm in peroxide-treated smectite D to relatively more octahedral Mg compared to peroxide-treated A and B. Other spectral features include a band at 2.24 μm, assigned to (Al,Fe³⁺)-OH, and a band at 2.40 μm commonly assigned to stretch-translation modes of (Fe³⁺,Mg)₂-OH (Gates et al., 2002; Frost et al., 2002) (Figure 7).

4. Discussion

4.1. Mechanisms of ferrous smectite oxidation

During oxidation in air-saturated solution, none of the ferrous smectites studied reached fully ferric compositions. The resistance of the ferrous smectites to complete oxidation after extended exposure to O₂ may relate to their trioctahedral structure. Iron oxidation in dioctahedral smectites has been proposed to be kinetically limited by cation migration required to maintain charge balance in the octahedral sheet (Manceau et al., 2000; Gorski et al., 2012). Trioctahedral smectites contain few octahedral vacancies to accommodate these structural rearrangements. Progressive Fe²⁺ oxidation in trioctahedral smectites may lead to positive charge buildup in the octahedral sheet. In the absence of charge redistribution by cation migration, each instance of Fe oxidation causes progressively stronger electrostatic repulsion and, therefore, increasing energetic unfavorability of further oxidation. Gorski et al. (2012) observed that in electrochemical reduction and reoxidation experiments, a significant fraction of octahedral Fe in Fe-rich smectites NAu-1 and NAu-2 was redox-inactive, presumably requiring extreme applied potentials to react. Similarly, a fraction of the Fe in the ferrous smectites of this study may not be labile under air-oxidizing conditions.

Another mechanism that may also explain incomplete oxidation is surface passivation. Since Fe³⁺ migration through the octahedral sheet is sluggish, Fe³⁺ forming at the edges of the smectite particles may form a passivating layer and limit further oxidation. Ilgen et al. (2012) attributed the observed inhibition of As³⁺ oxidation by the nontronite NAu-1, after initially fast reaction, to the passivation of the nontronite surface by formation of Fe³⁺oxide moieties.

In contrast to air-saturated solutions, peroxide solutions fully oxidize structural Fe in the ferrous smectites. Peroxide has a higher redox potential than O₂, such that Fe oxidation will proceed more readily even as positive charge buildup occurs in the octahedral sheet. Additionally, Fe²⁺ oxidation by peroxide proceeds by the Fenton reaction (Zepp et al., 1992), in which hydroxyl radical and superoxide radical are produced:

$$\text{F}{\text{e}}^{2+}+{\text{H}}_2{\text{O}}_2\to \text{F}{\text{e}}^{3+}+\text{HO}•+\text{O}{\text{H}}^{-}$$

$$\text{F}{\text{e}}^{3+}+{\text{H}}_2{\text{O}}_2\to \text{F}{\text{e}}^{2+}+\text{HOO}•+{\text{H}}^{+}$$

These radical species are powerful oxidants that may react with structural Fe²⁺, causing further oxidation.

Oxidation of trioctahedral, ferrous smectite was coupled with transformation to a dioctahedral structure. One mechanism by which octahedral site occupancy can decrease towards ⅔, as required for a dioctahedral smectite, is irreversible ejection of Fe³⁺ from the octahedral sheet (Farmer et al., 1971; Decarreau & Bonnin, 1986). Ferric iron ejected in this manner form nanoparticulate oxides and hydroxides that can accumulate in the interlayer, plate the surfaces of smectite particles, or develop into crystalline external phases (Farmer et al., 1971; Badaut et al., 1985; Decarreau & Bonnin, 1986; Treiman et al., 2014). In this study, all smectite compositions demonstrated Fe ejection upon peroxide treatment, but only smectites A and B also showed this effect with treatment in air-saturated solution, suggesting that more Fe-rich compositions are more susceptible to loss of structural Fe. One possible interpretation of this compositional dependency is that octahedral (Fe²⁺)₃-OH clusters are unstable after oxidation. The untreated ferrous smectites show no tendency towards Fe clustering, i.e., octahedral cations appear to have a random distribution among sites (Chemtob et al., 2015), so the abundance of (Fe²⁺)₃-OH clusters should be proportional to the initial octahedral Fe²⁺ content. The oxidation of these clusters to [(Fe³⁺)₃-OH]³⁺ would create local positive charge accumulation that might promote Fe ejection. In contrast, oxidation of isolated Fe²⁺ octahedra in a predominantly Mg-bearing octahedral sheet may not sufficiently destabilize the smectite structure to cause Fe ejection and iron oxyhydroxide formation.

4.2. Implications of VNIR reflectance spectra for smectite crystal chemistry

Michalski et al. (2015) recently characterized variations in the position of MM-OH and MMM-OH bands with smectite crystal chemistry. They determined that for ferric, Fe-rich dioctahedral smectites (molar Fe/Mg > 10), the MM-OH band occurs between 2.284 and 2.290 μm and that for trioctahedral smectites with Fe/Mg < 0.5 the MMM-OH position lies between 2.309 and 2.317 μm. For smectites with molar Fe/Mg ≈ 1, Michalski et al. observed a compositional dichotomy; trioctahedral smectites with that ratio had the MMM-OH peak near 2.31 μm, whereas di-trioctahedral smectites had the MM-OH peak near 2.290 μm.

The results of the Michalski et al. (2015) study can be compared to our air-reoxidized and peroxide-treated smectites, with the caveat that the Fe/Mg ratios in our oxidized smectites may be lower than in their unaltered counterparts because of Fe³⁺ exsolution and nanohematite formation. Our air-reoxidized and peroxide-treated smectites A and B (unaltered Fe/Mg = ∞ and 8.7, respectively) are dioctahedral and have MM-OH peaks at 2.284 and 2.286 μm, consistent with nontronites having molar Fe/Mg > 10 range as discussed by Michalski et al. (2015). For the dioctahedral smectite in peroxide-treated C and D (unaltered Fe/Mg = 1.7 and 0.5, respectively), the octahedral Fe³⁺-Mg MM-OH band occurs at longer wavelengths (2.295 and 2.299 μm, respectively). Our band positions occur at significantly shorter wavelengths compared to the range reported by Michalski et al. (2015) for dioctahedral smectites with Fe/Mg = 0–0.5 (2.310 to 2.316 μm) and at significantly longer wavelengths compared to their range for dioctahedral smectites with Fe/Mg = 1–5 (~2.290 μm). This apparent inconsistency may simply be a sampling problem, because Michalski et al. (2015) have only two analyses that overlap the Fe/Mg range for smectites C and D.

A majority (>70%) of Martian smectite detections by CRISM to date feature MM-OH bands near 2.30 μm (e.g., Carter et al., 2013; Ehlmann & Edwards, 2014; Michalski et al, 2015). Michalski et al. (2015) argue that this ubiquitous detection indicates abundant Martian smectites with Fe/Mg ratios between 5 and 10 which are elevated considerably with respect to unaltered Martian protolith (Fe/Mg ≈ 1.1 (McSween et al., 2009)). Michalski et al. (2015) suggest that such enrichment of Fe/Mg may require at least moderately oxic conditions at the time of basalt alteration. In terrestrial basaltic weathering profiles, atmospheric oxidation state is a critical control on Fe mobility (Rye & Holland, 1998; Babechuk et al., 2014). Under oxic conditions, Fe²⁺ is oxidized and retained in the weathering profile, causing Fe/Mg enrichment. In the absence of oxidants, Fe²⁺ and Mg²⁺ released in the weathering profile behave similarly and are minimally segregated. The crossover atmospheric oxygen level for Fe mobility depends on the rate of influx of weathering acids, but was estimated for terrestrial paleosols as ${\text{P}}_{{\text{O}}_2}\approx 0.0005$ atm by Rye & Holland (1998). Similarly, experimental dissolution of Fe-bearing olivine at ${\text{P}}_{{\text{O}}_2}<0.0001$ atm produced no fractionation of Fe from Mg in aqueous solution (Sugimori et al., 2012). During seafloor basalt weathering on Earth, other processes such as iron sulfide or oxy-hydroxide precipitation may fractionate smectite Fe/Mg from the protolith composition (Alt et al., 1986; Meunier et al., 2010), but such processes typically require high fluid fluxes and/or input of oxic altering seawater.

Our VNIR spectra suggest that an appeal to Noachian-era oxic conditions resulting in physical separation of Fe from Mg during basalt alteration is unnecessary. The partially oxidized products of our mixed Fe-Mg smectites C and D also have MM-OH band positions near 2.30 μm without physical separation of Fe and Mg. We suggest that CRISM spectral detections of MM-OH bands at 2.30 μm can equally well be explained by oxidation of trioctahedral ferrous smectite with crust-like (~1.1) or even more elevated Fe/Mg ratios to Fe-Mg dioctahedral smectite accompanied by ejection of Fe³⁺ as a ferric-bearing phase (e.g., nanoparticle hematite as in our study). Note that ejection of Fe³⁺ does not change the Fe/Mg ratio for the bulk sample, but does lower the Fe/Mg ratio for the dioctahedral smectite.

4.3. Implications for ferric smectite formation on Mars

Our studies demonstrate that solutions containing aqueous O₂ and H₂O₂, oxidants present in the Martian atmosphere and surficial regolith, can oxidize, over time, trioctahedral ferrous smectite to dioctahedral ferric smectite. Exposure to H₂O₂ aqueous solutions was structurally disruptive to all the smectites studied, ejecting Fe from the octahedral sheet and forming dioctahedral ferric smectite (nontronite) and secondary nanoparticulate hematite. Exposure of smectites to oxygen in air-saturated solutions resulted in nanoparticulate hematite formation for Fe²⁺-rich smectite compositions but not for mixed Fe-Mg smectites.

Oxidation in air-saturated solution produced fully ferric smectite only after a round of hydrothermal recrystallization, implying that structural relaxation and charge redistribution in the octahedral sheet may be required to produce nontronite from ferrous smectites. Can such structural relaxation occur in the Martian near-surface environment? On Earth, solution recrystallization of smectites occurs most readily in burial diagenesis environments at elevated pressure and temperature. Our experiments demonstrate that rapid recrystallization of smectites in aqueous solutions at 200° C can accommodate additional oxidation, but do not directly constrain the efficacy of this process at lower temperatures and water-rock ratios on geologic timescales. Two pieces of evidence suggest that these reactions can proceed near the arid Martian surface. First, smectite illitization, another structural transformation process that operates either in the solid-state or by dissolution/reprecipitation (Altaner & Ylagan, 1997), has been reported to occur pedogenically at slow rates in arid subaerial or near-surface environments (Singer, 1989; Bétard et al., 2009). Second, in this study, smectites inadvertently exposed to atmospheric O₂ at low concentrations developed VNIR reflectance spectral features consistent with transformation to nontronite. Although dry air oxidation was not an element of our initial experimental design, its effectiveness suggests that the sustained presence of fluids is unnecessary for Fe²⁺ oxidation and for smectite structural relaxation. Thus, given enough time, Martian near-surface trioctahedral ferrous smectites exposed to oxidants in the atmosphere and regolith likely also undergo structural relaxation towards dioctahedral structures.

VNIR reflectance spectra for the oxidized smectites are distinct from their ferrous trioctahedral precursors and resemble those for dioctahedral ferric smectites (nontronite). Ferrous smectite-bearing rock units later exposed to oxidizing conditions are thus a plausible explanation for widespread remote nontronite detections. Events triggering such a transformation could include 1) ancient exposure to an oxidizing atmosphere; 2) ancient exposure to oxidant-rich solutions like those proposed to have occurred at Meridiani Planum (Hurowitz et al., 2010) or Gale Crater (Rampe et al., 2017); or 3) extended, relatively modern exposure to the current atmosphere after exhumation. Pristine, unaltered ferrous smectites may only be observable by multispectral reflectance imaging at sites where impacts or mass wasting have freshly exposed such phases. The abundance of ferric smectites observed from orbit thus does not necessarily indicate the redox state of the early Martian surface, as oxidative processes would likely overprint any signatures of anoxic conditions in much of the exposed rock record.

4.4. Implications for the redox state of early Mars

Our experiments illustrate that detections of Martian ferric smectites in Noachian terrains need not imply oxidizing conditions during the Noachian because they may be postNoachian oxic alteration products of pre-existing ferrous smectites. Other orbital and rover-based mineralogic observations provide positive evidence for an early anoxic Mars. Near its landing site, the Curiosity rover at Gale Crater observed a gray-colored mudstone (Sheepbed Mudstone) below an oxidized surface layer (Grotzinger et al., 2014; Vaniman et al., 2014), dated to the Late Noachian / Early Hesperian (Grant et al., 2014). The Sheepbed Mudstone was found to contain, among other minerals, trioctahedral smectite and minor iron sulfide (Vaniman et al., 2014). Based on CheMin XRD data, the smectite in the Sheepbed Mudstone was previously interpreted as a ferrian (Fe³⁺ > Fe²⁺) saponite, largely because of equivalence in its (02l) d-spacing (4.59 Å) to the terrestrial Griffith saponite (Treiman et al., 2014). The (02l) d-spacing of our unaltered smectites C and D (4.60 and 4.58 Å, respectively; Table 3) are within error of the spacing observed for the Sheepbed Mudstone saponite (Vaniman et al., 2014; Chemtob et al., 2015), suggesting that ferrous compositions (Fe²⁺ > Fe³⁺) are also permitted within the constraints of the XRD data. The composition of the Martian saponite could not be calculated independently, but the aggregate chemistry of the saponite plus the amorphous component is more consistent with Mg-rich saponites like Griffith saponite and our two mixed Fe-Mg smectites C and D as opposed to our high-Fe ferrous smectites A and B (Vaniman et al., 2014). The minor amounts of iron sulfides that occur with the Martian smectite (Vaniman et al., 2014) should not persist under conditions of extensive iron oxidation and are perhaps remnants of previously more abundant sulfides. In Archean rocks on earth, detrital sulfides are interpreted as an indicator of anoxic conditions (e.g. Canfield et al., 2000). This work provides additional evidence that the CheMin analysis of the Sheepbed Mudstone is well explained by the initial formation of a Fe²⁺-Mg saponite by anoxic basalt alteration, with possible subsequent Fe²⁺ oxidation towards a ferrian composition.

Recent findings of chemical, mineralogical, and redox gradients in the Murray Formation, a mudstone-dominated sedimentary unit upsection from the Sheepbed Mudstone (Grotzinger et al., 2015), provide additional support for precursor ferrous smectites. XRD and evolved gas analyses of upper Murray Formation mudstones indicate that they contain hematite and both dioctahedral and trioctahedral smectites (Rampe et al., 2017; McAdam et al., 2017; Milliken et al., 2017; Bristow et al., 2017). Nontronite (a dioctahedral smectite) has also been detected by CRISM further upslope on Mt. Sharp. (Milliken et al., 2010; Fraeman et al., 2013; Poulet et al., 2014). This mineralogical shift from trioctahedral smectite-bearing downsection units has been interpreted as a Hesperian-aged environmental shift to drier, more oxidizing conditions (Rampe et al., 2017), or as coeval sediments deposited at different depths in a redox-stratified lake (Hurowitz et al., 2017). The assemblage of nontronite and hematite in the Murray Formation is striking because it mirrors the mineral assemblages observed in our experiments. Progressive oxidation of originally Fe²⁺-smectite-bearing deposits to nontronite, producing iron oxides as a byproduct of Fe ejection, is a potentially plausible mechanism for the Murray Formation mineralogical trend.

Orbital mineralogical datasets also provide evidence for an anoxic early Mars. CRISM has identified abundant and widespread Fe³⁺-bearing smectites in Noachian and early Hesperian terrains, but only rare and localized exposures of carbonates (Ehlmann & Edwards, 2014). Recent experimental evidence suggests that during aqueous oxic basalt alteration under a CO₂-rich atmosphere, Fe-Mg smectite formation is inhibited in favor of carbonate precipitation (Dehouck et al., 2016). Thus, the observed abundance of smectites and paucity of carbonates on Mars could imply poorly oxidizing conditions throughout the Noachian (Dehouck et al., 2016), although the observed relative mineral abundances could also be explained by phyllosilicate formation under mildly acidic conditions which inhibit carbonate stability (Peretyazhko et al., 2016).

CRISM multispectral analysis has identified stratigraphic sequences, at the late Noachian/early Hesperian-dated Mawrth Vallis, Nili Fossae, and other settings, in which Alrich phyllosilicate-bearing units overlie Fe-Mg phyllosilicate-bearing units. These sequences have been interpreted as mega-leaching profiles (Bishop et al., 2008; Ehlmann et al., 2009; Noe Dobrea et al., 2010; Gaudin et al., 2011; Zolotov & Mironenko, 2016). Although the Alrich phyllosilicate-bearing units at Nili Fossae have been proposed to contain iron oxides or oxyhydroxides (Gaudin et al., 2011), the Al phyllosilicate-rich units at Mawrth Vallis are apparently free of ferric phases, implying iron mobility during leaching. In terrestrial weathering profiles, Fe³⁺ is generally more mobile than aluminum, even under oxidizing conditions (Sheldon, 2003; Thomson et al., 2014); however, efficient fractionation of iron from aluminum occurs only in nonoxidizing weathering conditions (Rye & Holland, 1998; Babechuk et al., 2014). Therefore, Late Noachian Fe-Mg phyllosilicate-bearing units likely contained ferrous phases at the time of deposition and leaching, and conditions during leaching and Al-phyllosilicate formation were insufficiently oxidizing to convert Fe²⁺ to Fe³⁺.

If ferrous smectites were the initial phyllosilicate constituents of mid- and late-Noachian sedimentary units, anoxic conditions are required at the time of precipitation and deposition. The main driver of the oxidation of the Martian atmosphere is the imbalance between the introduction of reduced hydrogen by volcanic emissions and the non-reversible loss of hydrogen-bearing species (H, H₂, H⁺, H₂⁺) to space (Lammer et al., 2003). The identification of ferrous smectites in the Gale Crater deposits suggests that the rate of volcanic degassing at the Noachian-Hesperian boundary was sufficient to maintain anoxic conditions, overcoming both consumption of hydrogen by photochemically produced oxidants and hydrogen escape. Average rates of volcanic degassing have been estimated to increase from the late Noachian into the early Hesperian, then drop later in the Hesperian (Craddock & Greeley, 2009). The transition to global oxidizing conditions likely postdates the Yellowknife Bay deposits and may postdate the early Hesperian surge in volcanism. The abundant record of Fe-Mg smectites and the inferred presence of ferrous smectites in large leaching profiles at Mawrth Vallis are consistent with largely anoxic conditions through the late Noachian. The search for life on early Mars should thus in part focus on identifying mineralogical and geochemical signatures consistent with metabolisms that operate under anoxic conditions.

Key points:

• Oxidation of trioctahedral ferrous smectites by aqueous O₂ or H₂O₂ produces dioctahedral ferric smectites and Fe oxides or oxyhydroxides

• Oxidation produces systematic changes in VNIR reflectance spectra of ferrous iron-bearing smectites

• Orbitally detected Martian nontronite may be the product of oxidation of ferrous smectite precursors