Earth and Planetary Science Letters 430 (2015) 43–53

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

www.elsevier.com/locate/epsl

A redox-stratified ocean 3.2 billion years ago

Aaron M. Satkoski a,b,!, Nicolas J. Beukes c, Weiqiang Li d, Brian L. Beard a,b, Clark M. Johnson a,b

a University of Wisconsin-Madison, Department of Geoscience, 1215 West Dayton Street, Madison, WI 53706, United Statesb NASA Astrobiology Institute, United Statesc CIMERA, Department of Geology, University of Johannesburg, PO Box 524, Auckland Park, Johannesburg 2006, South Africad State Key Laboratory for Mineral Deposits Research, School of Earth Sciences and Engineering, Nanjing 210093, PR China

a r t i c l e i n f o a b s t r a c t

Article history:Received 7 April 2015Received in revised form 22 July 2015Accepted 10 August 2015Available online xxxxEditor: H. Stoll

Keywords:banded iron formationuraniumFe isotopes

Before the Great Oxidation Event (GOE) 2.4–2.2 billion years ago it has been traditionally thought that oceanic water columns were uniformly anoxic due to a lack of oxygen-producing microorganisms. Recently, however, it has been proposed that transient oxygenation of shallow seawater occurred between 2.8 and 3.0 billion years ago. Here, we present a novel combination of stable Fe and radiogenic U–Th–Pb isotope data that demonstrate significant oxygen contents in the shallow oceans at 3.2 Ga, based on analysis of the Manzimnyama Banded Iron Formation (BIF), Fig Tree Group, South Africa. This unit is exceptional in that proximal, shallow-water and distal, deep-water facies are preserved. When compared to the distal, deep-water facies, the proximal samples show elevated U concentrations and moderately positive !56Fe values, indicating vertical stratification in dissolved oxygen contents. Confirmation of oxidizing conditions using U abundances is robustly constrained using samples that have been closed to U and Pb mobility using U–Th–Pb geochronology. Although redox-sensitive elements have been commonly used in ancient rocks to infer redox conditions, post-depositional element mobility has been rarely tested, and U–Th–Pb geochronology can constrain open- or closed-system behavior. The U abundances and !56Fe values of the Manzimnyama BIF suggest the proximal, shallow-water samples record precipitation under stronger oxidizing conditions compared to the distal deeper-water facies, which in turn indicates the existence of a discrete redox boundary between deep and shallow ocean waters at this time; this work, therefore, documents the oldest known preserved marine redox gradient in the rock record. The relative enrichment of O2 in the upper water column is likely due to the existence of oxygen-producing microorganisms such as cyanobacteria. These results provide a new approach for identifying free oxygen in Earth’s ancient oceans, including confirming the age of redox proxies, and indicate that cyanobacteria evolved prior to 3.2 Ga.

! 2015 Elsevier B.V. All rights reserved.

1. Introduction

The timing of oxygenation of the world’s oceans and atmo-sphere has long been a topic of debate. A long-standing proposal embraced by many studies is that a step-wise oxygenation of the planet occurred, with the first rise in O2 taking place between 2.4 and 2.2 billion years ago (Ga), a time referred to as the Great Oxidation Event (GOE), and a second rise approximately 0.6 bil-lion years ago to nearly present-day levels (Holland, 1984). The world before the GOE is thought to have been largely anoxic, where O2 sinks out-paced O2 sources. Recent studies, however,

* Corresponding author at: University of Wisconsin-Madison, Department of Geo-science, 1215 West Dayton Street, Madison, WI 53706, United States.

E-mail address: satkoski@wisc.edu (A.M. Satkoski).

have called into question this strictly step-wise increase in plane-tary O2 levels, instead proposing transient increases in atmospheric oxygen before the GOE as far back as 3.0 Ga (Anbar et al., 2007;Crowe et al., 2013). Kasting (1992) predicted accumulation of O2in the shallow oceans, termed “oxygen oases”, before the GOE, but such environments would be limited to the photic zone through high biological primary productivity. Recent work has shown that oxygen oases could have been pervasive in the shallow ocean in the Archean under an essentially anoxic atmosphere (Olson et al., 2013). Geologic evidence for oxygen oases and a redox-stratified ocean has been di!cult to find, especially from the earlier parts of the Archean, although this is changing. Planavsky et al. (2014), for example, used Mo isotope data from iron formations in the Pon-gola Basin (Kapvaal Craton, South Africa) to infer free oxygen in the oceans at 2.95 Ga, and Riding et al. (2014) used REE variations in platform carbonates at Steep Rock (Superior Craton, Canada) to in-

http://dx.doi.org/10.1016/j.epsl.2015.08.0070012-821X/! 2015 Elsevier B.V. All rights reserved.

44 A.M. Satkoski et al. / Earth and Planetary Science Letters 430 (2015) 43–53

Fig. 1. Geologic map of the southwestern half of the Barberton greenstone belt showing the location of the BARB4 core drill site. Map is modified from Lowe (2013).

fer a redox-stratified ocean at 2.8 Ga. In contrast, using combined U–Pb and Fe isotope data on the 3.4 Ga Marble Bar Chert jasper (Pilbara Craton, Australia), Li et al. (2013a) argued that the oceans at that time were essentially anoxic. Although there is an emerg-ing view that intermittent rises in oxygen likely occurred prior to the GOE (Lyons et al., 2014), little is known about the potential transitions in oxygen contents in seawater that may have occurred over the 400 Myr between deposition the Marble Bar Chert and sedimentary rocks in the Pongola Basin.

Iron-rich chemical sedimentary rocks, including jaspers and Banded Iron Formations (BIFs) formed by oxidation of reduced iron, Fe(II), in an aqueous environment, and therefore have been prominent lithologies targeted for studies of the redox conditions of the oceans and atmosphere throughout the Archean and into the Proterozoic (Johnson et al., 2003; Rouxel et al., 2005; Planavsky et al., 2012; Li et al., 2013a). Iron isotope compositions of early Archean jaspers and BIFs are enriched in the heavy isotopes rela-tive to bulk crust (high 56Fe/54Fe ratios, or positive !56Fe values), which is interpreted to reflect partial oxidation of a reduced iron pool, indicating very low oxygen contents (Dauphas et al., 2004;Czaja et al., 2013; Li et al., 2013a). The concentrations of redox sensitive elements in BIFs, such as uranium (U), which is enriched in the oceans in its oxidized U(VI) form, have been used to argue for low O2 contents in the oceans prior to the GOE (Partin et al., 2013). A key issue that has been largely ignored, however, is the fidelity of redox proxies in terms of post-depositional mobilization through younger fluid–rock interactions, alteration, or metamor-phism. It is possible, for example, that inferences of high oxygen contents in the early Archean may reflect later fluid circulation un-der relatively oxidized (post-GOE) conditions; for example, using

238U–206Pb and 235U–207Pb geochronology, Li et al. (2012) showed that oxidation and U enrichment in the 3.4 Ga Apex Basalt (Pilbara Craton, Australia) occurred in the Phanerozoic and therefore does not provide a constraint on redox conditions in the early Archean as previously proposed. In this contribution, we present combined U–Th–Pb geochronology and stable Fe isotope results from the 3.23 Ga Manzimnyama BIF of the Fig Tree Group (Barberton green-stone belt, South Africa). Samples are from the BARB4 diamond-drilled scientific core, and include high-Fe and low-Fe cherts that record distinct depositional conditions which allow evaluation of redox conditions from deep- and shallow-ocean environments, re-spectively. Deep Archean seawater is generally thought to have been anoxic (Kamber et al., 2014), however, if oxygen was being produced by oxygenic photosynthesis, this production would likely happen in shallow water (photic zone) where cyanobacteria would thrive. Thus, the preservation of distinct deep- and shallow-water faces allows us to compare anoxic deep waters with potentially more oxygen enriched shallow waters. The unique combination of stable isotope compositions of redox-sensitive elements, and U–Th–Pb geochronology, allows us to distinguish primary signals from later fluid–rock interaction processes, which is critical for confidently inferring redox conditions in the oceans in the early Archean.

2. Geologic background and samples

The Manzimnyama BIF is part of the Fig Tree Group in the Barberton greenstone belt, South Africa (Fig. 1). The unit has also been described as a jasper or jasperlite, a term used for hematite–chert lithologies. The Barberton greenstone belt is a

A.M. Satkoski et al. / Earth and Planetary Science Letters 430 (2015) 43–53 45

volcanic-sedimentary sequence that ranges in age from "3.55 Ga to <3.23 Ga. The Fig Tree Group occurs between the volcanic Onverwacht Group and the siliciclastic Moodies Group (Fig. A.1). The Fig Tree Group is composed of greywacke, shale, chert, minor BIF/carbonate/barite, and felsic volcanic rocks. The age of the Fig Tree Group is well constrained by a spherule layer at 3258 ± 3 Ma(Byerly et al., 1996) at the base, and a volcanic unit (3226 ± 1 Ma) at the top (Kamo and Davis, 1994). The Fig Tree Group was de-posited in a variety of alluvial, fan delta, and deep- and shallow-water environments (Lowe, 2013). The metamorphic grade of the Barberton greenstone belt is relatively low for Early Archean ter-ranes, lower greenschist facies, and metamorphic temperatures could have reached "300 #C (Tice et al., 2004).

The Manzimnyama BIF is found near the base of the Fig Tree Group (Fig. A.1). Samples from this study are from the BARB4 sci-entific drill core, which was drilled though the Fig Tree Group in 2012. From the top of the core, the first "230 m are siliciclas-tic rocks, followed by "140 m of BIF that includes oxide- and carbonate-facies and minor siliciclastic layers, which is underlain by 80 m of missing/damaged core and black chert, chert breccia, and ultramafic volcanic rocks (Fig. A.2). The BIF sections primar-ily consist of laminated high-Fe chert, low-Fe chert, and siderite. Samples for this project are from the high- and low-Fe chert sam-ples (Figs. A.3 and A.4). Secondary veining is observed in parts of the core, and consists of silicate or calcite mineralogy (Fig. A.4). Sampling of the BIF avoided the localized silicate/carbonate veins, although a sample of a silicic vein was analyzed to understand the effects of secondary alteration on the primary chemistry and iso-topic compositions of the BIF. The high-Fe chert is fine grained, whereas the low-Fe chert samples have a granular texture (Figs. A.3 and A.4). The fine-grained high-Fe chert was likely deposited in a distal part of the basin in deep water, based on the finely lam-inated nature and lack of clastics. In contrast, the low-Fe samples are coarser in nature, consisting of sand-sized granules. Stefurak et al. (2014) report on silica granules from the Baberton Green-stone Belt, including those from Fe-rich and Fe-poor lithologies, and concluded that these formed via precipitation from the water column, including at shallow water depths. This would be anal-ogous to granular iron formations, which are interpreted to have formed in tidal zones (Simonson, 2003). The lack of compaction of the granules in the Manzimnyama BIF supports the interpretation that they formed in the water column, and following the mod-els of Stefurak et al. (2014) and Simonson (2003), are interpreted to have formed in a shallow-water, high-energy environment, in contrast to the finely laminated Fe-rich layers that would be gen-erally accepted to have accumulated below wave base at deeper water depths. Thus, we interpret the low-Fe granular samples to have precipitated in shallow waters nearer to the costal shelf and then transported into the deeper part of the basin where it was de-posited along with the deep-water high-Fe chert. Previous work on the high-Fe chert suggested that there was no evidence the Fe was secondary, and therefore was interpreted to be a primary chemical sediment (Hofmann, 2005).

3. Analytical methods

3.1. Sample preparation

Samples were cut directly from the core using a diamond blade. Rock chips were cleaned with acetone, 0.2 M HCL, and 18.2 M" H2O in an ultrasonic bath for 30 min to remove surface contamination before being dried and weighed. Samples ranged in weight from 100 to 250 mg. Samples were dissolved in a mixture of double-distilled HF and HCl in a capped Savellex beaker on a hotplate at "115 #C for several days. Samples were then heated in double-distilled 8M HCl to covert to chloride form, followed by

splitting into three aliquots: one for determining Pb isotope com-positions, one for determining Fe isotope compositions, and one for determining U–Th–Pb concentrations using isotope dilution.

3.2. Fe isotope and concentration measurements

Total iron concentrations were determined on a split of the Fe aliquot using the Ferrozine (with hydroxylamine hydrochloride) method after Stookey (1970). Based on replicate analyses and propagation of errors, Fe contents, as reported in wt.%, are esti-mated to have an uncertainty of 5%. For isotopic analysis, "100 µg of Fe from each sample were processed through anion-exchange resin using HCl (Beard et al., 2003). Purified Fe solutions were di-luted to 600 ppb and isotope measurements were made using a Micromass IsoProbe MC-ICP-MS using an Aridus® desolvating neb-ulizer. Solutions were aspirated at either 50 µL/min or 100 µL/min. Full details on mass spectrometry are reported in Beard et al.(2003). Isotope data are reported as 56Fe/54Fe using standard delta (!) notation in units of per mil (!) and are referenced to average igneous rocks (Beard et al., 2003):

!56Fe =!(56Fe/54Fe)sample/(

56Fe/54Fe)igneous rocks $ 1"% 1000

The external long-term reproducibility (2-SD) for !56Fe mea-surements, as determined from analyses of in-house Fe solutions is ±0.08!. Analysis of the IRMM-14 standard produced $0.08 ±0.08! (n = 329; 2-SD) on the igneous rock scale, which lies within error of the long-term value measured in our lab of $0.09!(Beard et al., 2003). Isotopic data are also reported in terms of 57Fe/54Fe in ! notation using an analogous definition (Table A.1), and all data fall along a mass-dependent relation. As a check on the ability of the chemical separation procedures and mass spec-trometry to recover an accurate Fe isotope composition, a synthetic Fe solution (Fe standard solution doped with matrix elements) was processed through ion-exchange chemistry and analyzed along-side samples, and the correct isotopic compositions were recovered within analytical uncertainty (Table A.1).

3.3. U–Th–Pb isotope and concentration measurements

The remaining two aliquots of dissolved sample were used for 1) Pb isotope measurements, and 2) U–Th–Pb concentration deter-minations by isotope dilution. The aliquot for concentration deter-minations was spiked with a mixed 235U–229Th–208Pb tracer solu-tion. Both aliquots were passed through anion-exchange resin us-ing 0.6 M HBr and Pb was eluted using 6M HCl. The initial column rinses for the spiked samples were collected for further purifica-tion of U and Th. The U–Th column split was processed through a second anion-exchange column using HNO3 and HCl, which pro-duced a mixed U–Th cut separated from matrix elements.

Lead isotope analysis was done using a Nu Instruments Nu Plasma II MC-ICP-MS using an Aridus® desolvating nebulizer. So-lutions were aspirated at either 50 µL/min or 100 µL/min. Samples were diluted to match the total ion intensity of the standards. In-strumental mass bias was corrected using thallium (205Tl/203Tl =2.389650) added to samples with a Pb/Tl ratio of "10 and then normalized to NBS-981. Based on the precision of standard anal-yses of USGS rock standard BCR-2, the uncertainty in 206Pb/204Pb, 207Pb/204Pb, and 208Pb/204Pb ratios is estimated to be 0.055% per amu (2-SD, n = 14, Table A.2).

Uranium and Th concentrations by isotope dilution were made using the concurrently measured 235U/238U and 229Th/232Th ra-tios, respectively, using a Micromass IsoProbe MC-ICP-MS and an Aridus® desolvating nebulizer. Samples were diluted to match the total ion intensity of the standards. Instrumental mass bias was corrected using a sample-standard bracketing method us-ing a mixture of an in-house Th standard (Ames Th) and U-500

46 A.M. Satkoski et al. / Earth and Planetary Science Letters 430 (2015) 43–53

(238U/235U = 1.00). Standard NBL-114 (238U/235U =137.88) was measured along side samples as a check on the sample-standard bracketing correction (Table A.2).

4. Results

4.1. Iron isotope and concentration results

The whole-rock Fe isotope and concentration data are reported in Table A.1. Iron concentrations (wt.%) have specific ranges that correlate with BIF texture, either granular or fine-grained (Figs. A.2 and A.3). The granular samples have Fe concentrations that range from 1.07 to 3.66 wt.%, whereas the fine-grained samples have Fe concentrations that range from 5.26 to 19.45 wt.%. The !56Fe val-ues of the fine-grained samples range from 0.84 to 0.29! and an average of 0.55!. In comparison, the granular (low-Fe) chert has a range of !56Fe values from 0.63 to 0.28!, and an average of 0.18!.

4.2. U–Th–Pb isotope results

The whole-rock U–Th–Pb isotope and concentration results are shown in Table A.2. Similar to the Fe isotope results, the U–Th–Pb results correlate with BIF texture. Lead concentrations for the low-Fe samples range from 0.78 to 0.23 ppm, whereas the high-Fe samples have Pb contents that range from 5.48 to 0.81 ppm. The U contents of the low-Fe samples range from 78.7 to 4.5 ppb, and 116.7 to 21.1 ppb for the high-Fe samples. Thorium contents in the low-Fe samples range from 9.0 to 2.1 ppb, whereas the high-Fe samples have a range of 141.6 to 21.9 ppb. 206Pb/204Pb ratios generally lie between 14.9 and 17.2, which is less radiogenic than modern average crust (206Pb/204Pb = 18.70; Stacey and Kramers, 1975), although one sample has an elevated 206Pb/204Pb ratio of 19.95). 207Pb/204Pb ratios range from 15.0 to 15.9, which overlaps that of average modern crust (207Pb/204Pb = 15.63; Stacey and Kramers, 1975). 208Pb/204Pb ratios vary from 33.1 to 34.6, all of which are significantly less radiogenic than modern average crust (208Pb/204Pb = 38.63; Stacey and Kramers, 1975).

5. Discussion

The goal of this work is to use redox sensitive elements in BIF to evaluate the O2 contents in 3.23 Ga seawater. In the fol-lowing discussion we approach this from two directions. First, we use Th contents to constrain the sedimentation rates of deep-and shallow-water BIF (Section 5.1), which then, in combination with Fe isotope data, we calculate the extent of Fe oxidation (Sec-tion 5.2). Second, we address post-depositional alteration of U (Section 5.3) before using it to constrain relative oxygen contents in deep- and shallow-water samples (Section 6). Both data sets give insights into ocean redox conditions, and in Section 7 we use these data to estimate the relative O2 content in 3.23 Ga seawa-ter as a function of water depth, as well as in comparison to older (3.4 Ga) jaspers.

5.1. Sedimentation rate derived from Th contents

The pH of Archean seawater is thought to have been slightly lower ("7.5) than modern (8.1), due to the increased concentration of CO2 in the atmosphere (Grotzinger and Kasting, 1993). Under these conditions the solubility of Th is extremely low in natural waters (Langmuir and Herman, 1980), therefore Th in these sam-ples must have been derived from aerosol or other clastic inputs and not from precipitation of dissolved Th. Even under very acidic conditions, where Th solubility increases, dissolved Th contents re-main extremely low. In terms of possible geographic variations in

Fig. 2. Cross plots of (A) Th versus core depth and (B) Fe concentration. In B, samples are divided into low-Fe and high-Fe samples. The low-Fe samples can be differ-entiated in core from the dark red, fine-grained high-Fe samples by their light pink color and a granular texture (Figs. A.3 and A.4). See Sections 5.1 and 5.2 for discussion on Fe/Th ratios and accumulation rate and extent of oxidation. (For in-terpretation of the colors in this figure, the reader is referred to the web version of this article.)

aerosol Th fluxes, Hsieh et al. (2011) showed that Th flux into the modern Atlantic Ocean does not vary greatly over a latitude and longitude range of "10 degrees. Considering all the samples stud-ied as part of this work are from a single core, the depositional area that these could represent, even considering any distance be-tween deep and shallow water samples, is likely to have been much smaller that 10 degrees of latitude/longitude. A constant Th flux into the basin is, also consistent with the lack of systematic variation in Th contents with depth in the core, only Fe contents (Fig. 2). Thus, thorium contents can be used as an indicator of relative sedimentation rates for the low-Fe and high-Fe cherts. As-suming a constant influx of Th, high Th samples are indicative of slower sedimentation rates, whereas low Th samples indicate a faster sedimentation rate. The average Th contents of the low-Fe chert samples are much less than those of the high-Fe chert sam-ples, which indicates that overall, the sedimentation rate for the low-Fe chert samples was faster than the high-Fe chert samples. The linear, positive correlations between Fe and Th (Fig. 2) also support this interpretation, demonstrating that total Fe contents are linked to the amount of time the sediment accumulated. We note that although the low-Fe samples represent clastic sedimenta-tion of previously precipitated chemical sediments, some of which have ooid-like structures, it would require extensive sediment sort-ing and preferential fractionation of the aerosol component to in-validate scaling sedimentation rates based on Th concentrations.

The measured Fe and Th abundances in the samples are inter-preted to reflect those of the primary sediment and unaffected by later metamorphism. Although the metamorphic history of the Bar-berton greenstone belt was such that greenschist-facies conditions were reached (Tice et al., 2004), the extremely low solubility of Th even at high temperatures and pressures (Baily and Vala Rag-narsdottir, 1994) indicates that Th should have remained immo-

A.M. Satkoski et al. / Earth and Planetary Science Letters 430 (2015) 43–53 47

bile during greenschist-facies metamorphism. Similarly, the solu-bility of hematite, the Fe-bearing phase analyzed in this study, is very low even under hydrothermal conditions at greenschist-facies temperatures (Diakonov et al., 1999), indicating that large-scale Fe mobility during metamorphism is unlikely. Moreover, because the Fe(III)aq-hematite fractionation factor remains near zero over a wide temperature range (Skulan et al., 2002), post-depositional fluid–rock interaction or metamorphism cannot have changed the measured Fe isotope compositions. Therefore, the correlated vari-ations in Fe and Th abundances are considered to reflect those of the primary sediment (Fig. 2). Such an interpretation is consistent with the arguments of Hoffman (2005) that Fe in Manzimnyama BIF is a primary feature.

5.2. Fe/Th–!56Fe relations

Iron in seawater is derived from continental weathering and oceanic hydrothermal fluids (Canfield et al., 2006), and, in the Archean, hydrothermal sources is thought to have dominated (e.g., (Derry and Jacobsen, 1990)). Stable Fe isotope studies indicate that hydrothermal Fe had !56Fe values between 0 and $0.5!(Yamaguchi et al., 2005). Oxidation of hydrothermal Fe(II) produces Fe(III) hydroxide/oxide precipitates, and complete or near-complete oxidation and precipitation will produce !56Fe values that are sim-ilar to that of the hydrothermal source (Beard and Johnson, 2004). In contrast, if partial oxidation of hydrothermal Fe(II) occurs, the !56Fe values of Fe(III) hydroxide/oxide precipitates will be signif-icantly higher than the hydrothermal sources, reflecting the fact that the equilibrium #56FeFe(OH)3–Fe(II) fractionation factor ranges from +2.5 to +4!, depending on Si/Fe ratios (Wu et al., 2012), although the increase in !56Fe values during oxidation may be mitigated by isotopic equilibration between aqueous Fe(II) and Fe(III) (Beard and Johnson, 2004). The !56Fe values of Fe(III) hy-droxide/oxide precipitates are therefore strongly dependent on the extent of oxidation, as modeled using a simple Rayleigh model or a more complex reaction-transport model (Li et al., 2013a). Here we use the Fe/Th ratio as a proxy for the extent of oxidation, based on the assumption that Th contents are a recorder of depositional rates (Section 5.1). Using Th abundances as a depositional clock, there are, however, differences in Fe/Th ratios for the low-Fe and high-Fe cherts that reflect differences in the extent of Fe precip-itation relative to Fe supply (Fig. 2). The average Fe/Th ratio of the high-Fe samples is 253 ± 64 (1-SD), significantly, and statis-tically, lower than the average Fe/Th ratio of the low-Fe samples (563 ± 212; 1-SD), indicating that the amount of Fe that precip-itated per unit time was higher in the low-Fe samples. This may seem counter-intuitive, however, but, as discussed in Section 5.1, the low Th contents of the low-Fe chert reflect relatively rapid deposition. The high Fe/Th ratios of the low-Fe chert, therefore, are interpreted to reflect a high extent of Fe oxidation and pre-cipitation, which is consistent with the inferred shallow water, near-shore depositional environment indicated by the sedimentol-ogy. In contrast, the low Fe/Th ratios of the high-Fe chert suggests limited Fe oxidation and precipitation per unit time, and this is consistent with the inferred deeper water setting and more limited extent of oxygen at deeper water depths. Two samples of high-Fe chert lie off the main Fe–Th trend on Fig. 2B, extending to higher Th contents, indicating a longer depositional period. The lack of a commensurate increase in Fe contents may simply reflect limi-tation of the concentration of dissolved Fe (ultimately limited by mineral solubility), which in turn would limit the maximum pos-sible Fe accumulation.

A powerful test of our model that the distinct Fe/Th ratios of the low-Fe and high-Fe cherts reflects differences in the extent of Fe oxidation and precipitation comes from Fe isotopes because the increase in !56Fe values that occurs during oxidation of aqueous

Fig. 3. Cross plot of Fe/Th and !56Fe for high- and low-Fe chert samples. The ob-served negative trend is interpreted to reflect increasing extent of oxidation with decreasing !56Fe values and increasing Fe/Th ratios. The range of Fe isotope com-positions for hydrothermal Fe(II) prior to oxidation is taken from Yamaguchi et al.(2005).

Fe(II) is proportional to the extent of oxidation and precipitation (Beard and Johnson, 2004; Li et al., 2013a), mitigated by any partial isotopic equilibration between aqueous Fe(II) and Fe(III) (Beard and Johnson, 2004). Small extents of oxidation of aqueous Fe(II) and precipitation of ferric oxides/hydroxides will produce the largest increases in !56Fe values, and our model predicts that these should be associated with the lowest Fe/Th ratios, which matches the data (Fig. 3). Larger extents of oxidation and precipitation should be associated with high Fe/Th ratios, and this too matches the data (Fig. 3). The relations in Fig. 3 provide compelling support for the interpretation that the low-Fe cherts reflect extensive oxidation of aqueous Fe(II) in a shallow marine environment, whereas the high-Fe cherts record lower extents of oxidation and precipitation in deeper water conditions. Samples that reflect 100% oxidation of aqueous Fe(II) would be those that have !56Fe values that match those of the initial fluid prior to oxidation, and hydrothermal Fe(II) sources should have had !56Fe values between $0.5 and 0.0!, likely closer to 0.0! in the Archean (Yamaguchi et al., 2005). From the trend observed in Fig. 3, we take the samples that have Fe/Th ratios of "1000 to reflect those that record complete or near-complete oxidation of aqueous Fe(II). This in turn allows us to scale the percent oxidation to Fe/Th ratio, and we assume this scales linearly to oxidation extent and therefore amount of Fe oxide pro-duction.

Scaling Fe/Th ratios to percent oxidation allows comparison of the !56Fe values with the reaction transport model of Li et al.(2013a), which demonstrates that, dependent upon the appropriate Fe(OH)3–Fe(II)aq fractionation factor, the Fe isotope variations are reasonably well described by variable extents of oxidation (Fig. 4). The model of Li et al. (2013a) assumes that oxidation occurs via downward diffusion of O2, and so greater extents of oxidation correlate with high O2 contents in seawater and shallower water depths. The Fe isotope results suggest limited oxidation for the deep-water high-Fe cherts, and significantly more extensive oxi-dation for the shallow-water low-Fe cherts. We note that some samples plot at lower !56Fe values than the model curves, par-ticularly those at low Fe/Th ratios, may reflect a smaller net Fe(III) oxide/hydroxide–Fe(II)aq fractionation factor, which, using the par-tial equilibration model of Beard and Johnson (2004) would corre-spond to an initially large kinetic isotope effect upon precipitation, and it is reasonable to assume such effects would be largest at low extents of oxidation and precipitation.

48 A.M. Satkoski et al. / Earth and Planetary Science Letters 430 (2015) 43–53

Fig. 4. Cross plot of percent oxidation, based on Fe/Th ratios, versus !56Fe. The blue and green bands are derived from modeling of the Fe isotope fractionation pro-duced during oxidation of aqueous Fe(II) using a reaction transport model (Li et al., 2013a), shown for two different #56FeFe(OH)3–Fe(II) fractionation factors of 4!and 2.5! (Wu et al., 2012). The band widths reflect two distinct initial hydrother-mal Fe(II) sources of !56Fe = 0 ! and $0.5!. Oxidation in the reaction transport model is accomplished by O2 produced through oxygenic photosynthesis that dif-fuses downward, interacting with aqueous Fe(II) that ascends from deeper water levels. Oxidation percent for the BIF samples is calculated by linearly scaling Fe/Th ratios to percent oxidation (see Fig. 3). Note that one anomalously high-Fe/Th sam-ple is omitted (see Fig. 3). Samples that plot to the left of the model curves, at low Fe/Th ratios, could reflect a lower net oxide-aqueous Fe isotope fractionation factor, as may be produced during partial equilibration between aqueous Fe(III) and Fe(II), as well as a kinetic isotope effect during early stages of precipitation (Beard and Johnson, 2004). (For interpretation of the colors in this figure, the reader is referred to the web version of this article.)

5.3. U–Th–Pb isotopes: evaluation of closed-system behavior

If the oxidizing potential of shallow ocean water at 3.2 Ga was greater than that of deep water, as suggested by the Fe/Th–!56Fe relations, the concentration of other redox-sensitive elements should also vary with water depth, and here we focus on U abundances. Before we interpret the U data in the context of potential seawater concentrations, however, we test the possi-bility that post-depositional element mobility occurred through metamorphism and fluid–rock interaction using the U–Th–Pb iso-tope system, which can rigorously constrain open- or closed-system behavior since deposition at 3.2 Ga. Although many stud-ies have been published on redox-sensitive elements as a proxy for redox conditions in the Precambrian, such studies generally do not address post-depositional element mobility. We argue that this is particularly important to do for redox-sensitive el-ements that could be mobilized under oxidizing conditions, be-cause it could be possible, for example, to have a U-enriched sample that simply reflects mobilization during fluid–rock in-teraction after Earth’s atmosphere became oxidized (Li et al., 2012). Closed- or open-system behavior for U and Pb may be evaluated using 238U/204Pb–206Pb/204Pb, 235U/204Pb–207Pb/204Pb, 232Th/204Pb–208Pb/204Pb, and 207Pb/204Pb–206Pb/204Pb isochron di-agrams, provided the depositional age and initial Pb isotope com-positions are well known. As noted above in Section 2, the de-positional age of the Manzimnyama BIF is well constrained by a 3.258 Ga spherule layer and a 3.225 Ga volcanic unit, and we as-sume an age of 3.23 Ga. The possible range in initial Pb isotope compositions of the Manzimnyama BIF, which would reflect con-tributions from aerosol, clastic, and authigenic (seawater) compo-nents, is assessed using Pb isotope compositions for igneous rocks

Fig. 5. (A) Plot of 238U/204Pb versus 206Pb/204Pb for Manzimnyama BIF samples and a silicic vein sample. The Barberton common Pb composition is from a 3.23 Ga con-formable galena deposit (Stacey and Kramers, 1975), which is used to tie down the initial Pb for the 3.23 Ga reference isochron. Based on the relations in Fig. A.5, how-ever, initial 206Pb/204Pb ratios for the Manzimnyama BIF could have been higher or lower than the “Barberton common Pb”, and the blue box around the 3.23 Ga isochron reflects this range. Samples that are considered closed, or “nearly closed”, with respect to U and Pb are marked. Many samples, however, fall to the left of the 3.23 Ga isochron field, and are interpreted to reflect Pb addition at "2.0 Ga. A 2.0 Ga reference isochron is shown, whose initial is tied by the Stacey–Kramers crustal value at 2.0 Ga; this closely matches the Pb isotope composition of the sili-cic vein. Estimates for the possible range in Pb isotope compositions of the 2.0 Ga Pb addition are shown by the grey box around the 2.0 Ga reference isochron, as calculated assuming a range of µ values (238U/204Pb) between µ = 14 and 6, de-rived from the Barberton common Pb composition at 3.23 Ga. (B) Plot of 235U/204Pb versus 207Pb/204Pb for BIF samples and a silicic vein sample. The 3.23 Ga reference isochron, and range in initial Pb isotope compositions, follow that described in (A). The 2.0 Ga reference isochron, and range in initial Pb, also follows that described in (A). (For interpretation of the colors in this figure, the reader is referred to the web version of this article.)

and ore Pb from the Barberton greenstone belt. A conformable galena deposit at 3.23 Ga (Stacey and Kramers, 1975) provides one assessment of initial Pb isotope compositions, and we define this as “Barberton common Pb”. Data from intrusive rocks, however, suggest a wider range in initial Pb isotope compositions, partic-ularly for 207Pb/204Pb. A full discussion of the range in initial Pb isotope compositions used in our assessment of element mobility is given in the Supplementary Materials.

In Figs. 5A and 5B the 238U/204Pb–206Pb/204Pb and 235U/204Pb–207Pb/204Pb relations are shown for the Manzimnyama BIF, along with reference 3.23 Ga isochrons tied to the range of possi-ble initial 206Pb/204Pb (12.30 to 12.78) and 207Pb/204Pb (13.97 to 14.33) ratios in the Barberton greenstone belt (Supplementary Ma-terials). BIF samples that plot within the 3.23 Ga isochron field are taken to have remained closed in terms of U and Pb since deposition. Note that the wider isochron field in Fig. 5B simply re-flects the wider range in possible initial 207Pb/204Pb ratios at the time of deposition, as reflected in the wide range in 207Pb/204Pb ratios for Barberton plutons (Fig. A.5). Four low-Fe samples plot within the 3.23 Ga isochron field for 238U/204Pb–206Pb/204Pb and 235U/204Pb–207Pb/204Pb, and thus are considered isotopically closed. One high-Fe sample also plots within the isochron field

A.M. Satkoski et al. / Earth and Planetary Science Letters 430 (2015) 43–53 49

and hence is considered closed. One low-Fe sample that has a very high 206Pb/204Pb ratio (19.95) plots slightly to the right of the isochron field (equivalent to 5% U enrichment or Pb loss), and this sample is considered “nearly closed”. We note that the criterion we use for closed-system behavior, where both the age and initial Pb isotope composition match that of geologic constraints, is the basis for all isochron-based geochronology.

The remaining Manzimnyama BIF samples plot to the left of the 3.23 Ga isochron field in Figs. 5A and 5B, which can be ex-plained by U loss, Pb addition, or a combination of the two. The samples that plot furthest to the left of the 238U/204Pb–206Pb/204Pb and 235U/204Pb-207Pb/204Pb isochrons tend to have the highest Pb contents, particularly the high-Fe cherts (Table A.2), suggest-ing that most samples that are not closed experienced Pb addi-tion. Evidence for Pb addition is most clearly indicated by the 232Th/204Pb–208Pb/204Pb relations (Fig. A.7). Because Th contents of the Manzimnyama BIF are extremely low, the thorogenic Pb in the samples is easily overprinted by very small amounts of post-depositional Pb addition. This is illustrated by the fact that the six samples that are considered closed or nearly closed for U–Pb plot far to the left of the 232Th/204Pb–208Pb/204Pb isochron (Fig. A.7). It is critical to recognize, however, that the relatively high initial U contents of the Manzimnyama BIF, relative to Th contents, means that uranogenic Pb is much more di!cult to overprint during post-depositional Pb addition than thorogenic Pb, and hence closed- or open-system behavior is well constrained for U–Pb.

207Pb/204Pb–206Pb/204Pb variations for the Manzimnyama BIF samples (Fig. 6) provide strong constraints on the age of the Pb addition event, as well as possible recent U addition. Samples that define the left-most edge of the 207Pb/204Pb–206Pb/204Pb BIF ar-ray (Fig. 6) are coincident with a Pb–Pb chord that is produced via crustal evolution of variable U/Pb ratios between 3.23 and 2.0 Ga that is tied to the Barberton common Pb initial. All of the BIF samples that plot along the 2.0 Ga Pb–Pb chord have very low U/Pb ratios, indicating that common Pb dominates their Pb iso-tope compositions. These relations provide strong evidence that Pb addition occurred at "2.0 Ga. In addition, the silicic vein sam-ple plots very close to the 2.0 Ga Pb–Pb chord, indicating that it is likely that veining occurred at "2.0 Ga and was associated with the Pb addition event. The evidence that the "2.0 Ga Pb addition event is associated with veining that is laterally hetero-geneous indicates that the extent of Pb addition should be variable depending on the samples. It is also important to note that the samples considered closed come form the lower part of the core, which is furthest from sections of the core that contain the high-est density of veining. A test of the conclusions on closed- or open-system behavior based on U–Pb isochron diagrams (Fig. 5) comes from 207Pb/204Pb–206Pb/204Pb relations (Fig. 6). The one sample classified as “nearly closed” for U–Pb, which has the high-est 206Pb/204Pb ratio, plots within the 3.23 Ga isochron array for 207Pb/204Pb–206Pb/204Pb, suggesting that the small amount of U addition ("5%, not accounting for U decay) that is suggested in Fig. 5 probably reflects recent U addition.

There is abundant evidence for a "2.0 Ga event in the Barber-ton greenstone belt. Although peak metamorphism reached lower greenschist facies in the Archean ("300 #C; Tice et al., 2004), 40Ar/39Ar age spectra on Fig Tree Group sediments, as well as 87Rb–87Sr geochronology on Onverwacht Group cherts, indicate a hydrothermal fluid-flow event of "2.1–2.0 Ga age (Weist and Wasserburg, 1987; de Ronde et al., 1991). In both cases, the source of hydrothermal fluid migration is considered to be the "2.1–2.0 Ga Bushveld large igneous province, and we propose that this was the driver of the Pb addition event seen in the Manz-imnyama BIF samples. There is no evidence, however, that the "2.0 Ga hydrothermal event mobilized Fe, based on the correla-tion between Fe/Th and !56Fe. No dissolution features are visible

Fig. 6. (A) 207Pb/204Pb versus 206Pb/204Pb diagram (“Pb–Pb”) showing Pb isotope variations for the low- and high-Fe chert samples and the silicic vein sample. Fig-ure B is a plot of the data shown from the dashed box in A. The terrestrial Pb evolution curve is from Stacey and Kramers (1975). The Barberton common Pb is from a 3.23 Ga conformable galena deposit (Stacey and Kramers, 1975). The range in possible Pb compositions at 3.23 Ga is shown by the blue box. The chert sam-ples that define the left-most end of the array (see “B”) fall along a 2.0 Ga Pb–Pb chord that is obtained through evolution of various µ values (238U/204Pb) from the Stacey–Kramers curve at 3.23 Ga. Solid black circles mark compositions for µ = 6and 14, and the chord between these points likely represents that possible range of common Pb isotope compositions that were involved in the "2.0 Ga Pb addition event; these compositions form the upper and lower bounds to the 2.0 Ga reference isochrons in Figs. 5A and 5B. The coincidence of the left side of the chert data array and the 2.0 Ga Pb evolution provides strong evidence that the Pb addition event is "2.0 Ga. (B) Detailed plot of data shown in the dashed box in A. (For interpretation of the colors in this figure, the reader is referred to the web version of this article.)

in the core and alteration appears to be localized to the secondary veins (Fig. A.3). Further evidence that Fe isotopes are not affected by hydrothermal fluids comes from the recent work of Li et al.(2013b) in which they showed that the 2.5 Ga BIF from the Dales Gorge in Australia had undergone hydrothermal alteration, but the Fe isotopes were unaffected and reflected primary values. In ad-dition, post-GOE fluid flow (which is localized and not pervasive throughout the BARB4 core) would likely involve oxidized fluids, and such fluids could not have mobilized significant quantities of Fe(III) (Diakonov et al., 1999).

6. U concentration in seawater

The only significant source of U in seawater is from oxida-tive weathering of uraninite, which is most abundant in evolved (“granitic”) continental crustal rocks (Partin et al., 2013). Recent work has shown that, during the Archean, uraninite would not have been stable in fluids that had higher than 40 nM O2 (John-son et al., 2014). Values much larger than this ("10 µM) have been estimated based on geochemical studies of the 2.8 Ga Steep Rock Group, Canada (Riding et al., 2014). In addition, work by

50 A.M. Satkoski et al. / Earth and Planetary Science Letters 430 (2015) 43–53

Fig. 7. Cross plot of Fe2O3 and U comparing the deep/shallow water Manzimnyama samples from this study. Use of the U!/Fe2O3 ratio allows normalization of U abundances as a function of Fe contents, and from this the U in seawater (Useawater) can be calculated using appropriate distribution coe!cients. Shown for reference is the concentration of U in modern seawater, calculated using the same absorption coe!cient as applied to the Archean samples. Using only samples that have remained closed or nearly closed for U–Pb (see text for discussion), the deep-water samples indicate a Useawater concentration of "0.04 ppb, whereas the shallow water samples indicate a Useawaterconcentration of "0.14 ppb, suggesting that a discrete redoxcline existed between deep and shallow waters at 3.23 Ga. Comparison between the distal, deep-water samples from this study and those from the 3.46 Ga Marble Bar Chert (green line; where ocean water was interpreted to be anoxic), show similar Useawater concentrations. This implies that a redoxcline formed sometime between 3.46 and 3.23 Ga. Uranium concentration in modern seawater is from Barnes and Cochran (1990). SW = seawater. (For interpretation of the colors in this figure, the reader is referred to the web version of this article.)

Kasting (1992) suggests that O2 contents could reach "10 µM un-der an atmosphere that would still be considered anoxic. In urani-nite, U exists as U(IV), which is insoluble in aqueous solutions, al-though oxidative weathering will produce soluble U(VI) (Langmuir, 1978), which in turn would enter the ocean via riverine input. Ex-perimental work has shown that the absorption of U onto Fe(III) oxides/hydroxides is "100 times more e!cient than absorption on to silica (Ames et al., 1983). Therefore, Fe(III) oxides/hydroxides are likely the dominant control on U sorption in chemically precipi-tated rocks. From this, uranium in chemically precipitated rocks, such as BIFs can be used to calculate the uranium concentration in the water from which it precipitated by normalizing U to Fe contents (U/Fe2O3 ratio) and using experimentally determined ab-sorption coe!cients for U(VI) sorption to Fe(III) oxides/hydroxides.

Following Li et al. (2013a), calculating the U concentration of ancient ocean water relies on the following equation:

Umeasured = Uabsorbed + Udetrital + Uadded/lost $ Udecay (1)

Uabsorbed is the U that was absorbed onto primary Fe(III) hydrox-ides in 3.23 Ga ocean water. The Udetrital component is the U that was added to the BIF due to siliciclastic contamination, which, in the low-clastic setting of the Manzimnyama BIF, was likely domi-nated by aerosol particles. The Udetrital component can be robustly accounted for by using the measured Th contents and assuming an average crustal U/Th ratio of 0.25; support for this ratio comes from data from Barberton plutons shown in Fig. A.6 which scatter about an average U/Th ratio of "0.25. The Udecay component is ac-counted for by calculating the amount of 235U and 238U that has decayed over the last 3.23 billion years; this correction is most ac-curate for samples that have been demonstrated to have remained closed with respect to U–Pb since formation 3.23 by ago (Fig. 5). Using the decay constants for 235U and 238U and an age of 3.23 billion years, the amount of U lost to decay is 44%. The Uadded/lostcomponent is the U that was added or lost after deposition of the Manzimnyama BIF. For the purposes of this study we will only con-sider samples that have remained closed with respect to U–Pb, and so this component can be ignored. It is important to note, however, that all evidence suggests that open-system behavior for U–Pb ap-pears to be largely restricted to Pb addition, rather than U mobility, and so it is likely that many of the non-closed samples would yield

Fig. 8. Correlation between O2 content in the photic zone and the resultant !56Fe value of Fe(III)-oxides/hydroxides. Modeling parameters same as in Fig. 4, and fol-lows the approach outlined in Li et al. (2013a). Uranium in seawater (Useawater) contents are shown from Fig. 7. For comparison, data from the 3.4 Ga Marble Bar Chert are shown, whose very high !56Fe values provide strong constraints of very low O2 contents, <0.0001 µM (Li et al., 2013a). The moderately positive !56Fe val-ues of the high-Fe chert in this study suggest O2 contents between 0.0001 and 0.001 µM, although the non-linear nature of the !56Fe–O2 curve permits higher O2contents. The relatively low !56Fe values of the low-Fe chert provide qualitative evi-dence for higher O2 contents, although the sensitivity of the model becomes low as the !56Fe value of the source hydrothermal Fe(II) is approached (see discussion in Li et al., 2013a); maximum O2 contents of 10 µM are assumed (Olson et al., 2013).

accurate U contents for 3.23 Ga oxides. Although we cannot rigor-ously prove that to be true, experimental work has shown that once U is incorporated into iron oxides, it is resistant to extraction and re-mobilization (Massey et al., 2014 and references therein).

To calculate the concentration of U in seawater, we solve equa-tion (1) for Uabsorbed. The amount of U that was absorbed on the primary iron oxide precipitates at 3.23 Ga is here termed U! , and is defined by the equation:

A.M. Satkoski et al. / Earth and Planetary Science Letters 430 (2015) 43–53 51

Fig. 9. A conceptual cross-section showing the depositional environment of the Manzimnyama BIF. The distal, deep-water Fe-rich BIF samples were oxidized in deeper water where O2 contents were lower than shallow-water Fe-poor BIF samples that were precipitated entirely above the redoxcline, where O2 contents were uniformly elevated. The low-Fe samples are granular (Fig. A.3), and they are found in slump and debris features, which supports the interpretation that they record precipitation in a shallow-water, near-shore environment, followed by transportation to a deeper part of the basin and deposited along with the more distal, deeper water (high-Fe) samples. See Section 6for discussion of U! . The range for !56Fe values of hydrothermal fluids is discussed in Yamaguchi et al. (2005).

U! = Umeasured + Udecay $ Udetrital (2)

The U! value in equation (2) can be directly related to the U con-centration in seawater by considering the Fe contents and absorp-tion coe!cient of U onto oxides/hydroxides of 104–106 (Wazne et al., 2003; Davis et al., 2004). The U seawater concentrations (Useawater) reported here are calculated using a conservative 104absorption coe!cient. It is important to note that because both the high- and low-Fe samples are calculated using the same ab-sorption coe!cient, the absolute value of absorption coe!cient used is not as important as is the relative difference in Useawaterbetween the two data sets, which provides a relative estimate of the Useawater values for shallow- and deep-water. Calculated using the four shallow-water (low-Fe) samples that remained closed for U–Pb, Useawater concentrations are estimated to have ranged from 0.08 to 0.19 ppb, and we assume an average value for the shallow ocean at 3.23 Ga of 0.14 ppb. This range is in stark contrast to the deeper water (high-Fe) sample that has remained closed for U–Pb, which produces a Useawater concentration of 0.04 ppb (Fig. 7). For comparison, Li et al. (2013a) estimated a Useawater concentration of 0.02 ppb at 3.4 Ga based on analysis of the Marble Bar Chert (Pilbara Craton, Australia), calculated using the same 104 absorp-tion coe!cient. Although there is uncertainty in the appropriate U absorption coe!cient to use in the calculations, the fact that we have used the same distribution coe!cient to compare data means the relative differences seen in the deep- and shallow-water chert of this study are significant, and comparison to the results from the Marble Bar Chert is robust. Collectively, these results suggest that the oceans were essentially anoxic at 3.4 Ga, but by 3.2 Ga the shallow ocean became significantly enriched in O2 such that dissolved U contents rose significantly, but deep-water settings re-mained nearly anoxic. Although we have focused our calculations on the samples that are closed for U–Pb, the contrast in U contents inferred from the deep- and shallow-water cherts holds if the en-tire data set is considered (Fig. 7).

7. Estimation of O2 contents in 3.23 Ga seawater

Using a reaction-transport model, an estimate may be made for the O2 contents in the shallow ocean based on the !56Fe values for iron oxides (Li et al., 2013a). It is not possible to quantify O2 contents based on U abundances because U con-tents are not limited by mineral solubility in modern seawater

(Klinkhammer and Palmer, 1991), and this seems likely to have been the case in the Archean. Use of !56Fe values for iron oxides to constrain ocean water O2 contents suggests that the high-Fe chert samples could have formed from seawater that had a maximum O2content of 0.4 µM, whereas the low-Fe chert samples could have formed from seawater that had a maximum O2 content of "10 µM (Fig. 9). The "10 µM O2 content is the maximum predicted dur-ing the Archean based on modeling possible oxygen levels in the shallow oceans beneath an anoxic atmosphere (Olsen et al., 2013). Although it is di!cult to confidently constrain O2 contents using !56Fe values for iron oxides as the isotopic composition approaches that of the hydrothermal source, the difference in average !56Fe values from the shallow-water (low-Fe) and deep-water (high-Fe) cherts is consistent with the relative redox state inferred from the U abundances. Moreover, the !56Fe–O2 modeling indicates that O2contents at 3.23 Ga were significantly higher than that inferred for 3.46 Ga based on the Marble Bar Chert (Fig. 8).

In Fig. 9 we illustrate a conceptual model for redox zona-tion in the oceans as recorded by the 3.23 Ga Manzimnyama BIF. Although oxidation of hydrothermally sourced Fe(II) can be ac-complished through anaerobic photoautotrophic microorganisms, which use aqueous Fe(II) as an electron donor (Widdel et al., 1993), or by free oxygen, the elevated U abundances we cal-culate for the shallow oceans argues the oxidation mechanism was through free O2. Some oxidative weathering on the conti-nents must have occurred to transport soluble U(VI) to the oceans, although such weathering may have been quite low, and may have limited U(VI) availability. The co-variation in Useawater and !56Fe values between proximal (low-Fe) and distal (high-Fe) sam-ples indicates that oxygen was enriched in the photic zone. We propose that the proximal (low-Fe) samples precipitated largely above the redoxcline in shallow water, whereas the distal (high-Fe) samples precipitated from a larger portion of the water col-umn that included the upper, more oxygen-rich zone, but also lower parts of the water column where oxygen contents would be lower (Fig. 9). Based on comparison with results from the 3.46 Ga Marble Bar Chert (MBC), Australia, a redoxcline formed in seawa-ter after 3.46 Ga but before 3.23 Ga. Oxygenic photosynthesis by cyanobacteria has been suggested as a mechanism for producing O2 at 2.8 Ga (Riding et al., 2014) and 3.0 Ga (Crowe et al., 2013;Planavsky et al., 2014), and we suggest the same as a mechanism for O2 generation here. The work presented here, therefore, pushes

52 A.M. Satkoski et al. / Earth and Planetary Science Letters 430 (2015) 43–53

back the earliest evidence for oxygenic photosynthesis by "200 million years.

Acknowledgements

This study was supported by the NASA Astrobiology Institute at the University of Wisconsin-Madison and the DST-NRF Centre of Excellence for Energy and Mineral Resource Analysis (CIMERA) at the University of Johannesburg. Comments by an anonymous reviewer, Matthew Fantle, and the editor, helped to improve the manuscript.

Appendix A. Supplementary material

Supplementary material related to this article can be found on-line at http://dx.doi.org/10.1016/j.epsl.2015.08.007.

References

Ames, L.L., McGarrah, J.E., Walker, B.A., 1983. Sorption of trace constituents from aqueous solutions onto secondary minerals. I. Uranium. Clays Clay Miner. 31, 321–334.

Anbar, A.D., Duan, Y., Lyons, T.W., Arnold, G.L., Kendall, B., Creaser, R.A., Kaufman, A.J., Gordon, G.W., Scott, C., Garvin, J., Buick, R., 2007. A whiff of oxygen before the Great Oxidation Event? Science 317, 1903–1906.

Baily, E.H., Vala Ragnarsdottir, K., 1994. Uranium and thorium solubilities in subduc-tion zone fluids. Earth Planet. Sci. Lett. 124, 119–129.

Barnes, C.E., Cochran, J.K., 1990. Uranium removal in oceanic sediments and the oceanic U balance. Earth Planet. Sci. Lett. 97, 94–101.

Beard, B.L., Johnson, C.M., 2004. Fe isotope variations in the modern and ancient Earth and other planetary bodies. In: Johnson, C.M., Beard, B.L., Albarède, F. (Eds.). Geochemistry of Non-Traditional Stable Isotopes, Rev. in Mineralogy and Geochemistry, vol. 55. Mineralogical Society of America and Geochemical Soci-ety. Washington DC, pp. 319–357.

Beard, B.L., Johnson, C.M., Skulan, J.L., Nealson, K.H., Cox, L., Sun, H., 2003. Applica-tion of Fe isotopes to tracing the geochemical and biological cycling of Fe. Chem. Geol. 195, 87–117.

Byerly, G.R., Kröner, A., Lowe, D.R., Todt, W., Walsh, M.M., 1996. Prolonged magma-tism and time constraints for sediment deposition in the early Archean Barber-ton greenstone belt: evidence from the upper Onverwacht and Fig Tree groups. Precambrian Res. 78, 125–138.

Canfield, D.E., Rosing, M.T., Bjerrum, C., 2006. Early anaerobic metabolisms. Philos. Trans. R. Soc. B 361, 1819–1836.

Crowe, S.A., Døssing, L.N., Beukes, N.J., Bau, M., Kruger, S.T., Frei, R., Canfield, D.E., 2013. Atmospheric oxygenation three billion years ago. Nature 501, 535–539.

Czaja, A.D., Johnson, C.M., Beard, B.L., Roden, E.E., Li, W., Moorbath, S., 2013. Bio-logical Fe oxidation controlled deposition of banded iron formation in the ca. 3770 Ma Isua Supracrustal Belt (West Greenland). Earth Planet. Sci. Lett. 363, 192–203.

Dauphas, N., van Zuilen, M., Wadhwa, M., Davis, A.M., Marty, B., Janney, P.E., 2004. Clues from iron isotope variations on the origin of early Archaean banded iron formations from Greenland. Science 306, 2077–2080.

Davis, J.A., Meese, D.E., Kohler, M., Curtin, G.P., 2004. Approaches to surface com-plexation modeling of uranium(VI) absorption on aquifer sediments. Geochim. Cosmochim. Acta 68, 3621–3641.

de Ronde, C.E.J., Hall, C.M., York, D., Spooner, E.T.C., 1991. Laser step-heating 40Ar/39Ar age spectra from early Archean ("3.5 Ga) Barberton greenstone belt sediments: a technique for detecting cryptic tectono-thermal events. Geochim. Cosmochim. Acta 55, 1933–1951.

Derry, L.A., Jacobsen, S.B., 1990. The chemical evolution of Precambrian seawater: evidence from REEs in banded iron formations. Geochim. Cosmochim. Acta 54, 2965–2977.

Diakonov, I.I., Schott, J., Martin, F., Harrichourry, J.C., Escalier, J., 1999. Iron(III) solubility and speciation in aqueous solutions. Experimental study and mod-eling: Part 1. Hematite solubility from 60 to 300 #C in NaOH–NaCl solutions and thermodynamic properties of Fe(OH)4-(aq). Geochim. Cosmochim. Acta 63, 2247–2261.

Grotzinger, J.P., Kasting, J.F., 1993. New constraints on Precambrian ocean composi-tion. J. Geol. 101, 235–243.

Hofmann, A., 2005. The geochemistry of sedimentary rocks from the Fig Tree Group, Barberton greenstone belt: implications for tectonic, hydrothermal and surface processes during mid-Archaean times. Precambrian Res. 143, 23–49.

Holland, H.D., 1984. The Chemical Evolution of the Atmosphere and Oceans. Prince-ton University Press.

Hsieh, Y., Henderson, G.M., Thomas, A.L., 2011. Combining seawater 232Th and 230Th concentrations to determine dust fluxes to the surface ocean. Earth Planet. Sci. Lett. 312, 280–290.

Johnson, C.M., Beard, B.L., Beukes, N.J., Klein, C., O’Leary, J.M., 2003. Ancient geo-chemical cycling in the Earth as inferred from Fe isotope studies of banded iron formations from the Transvaal Craton. Contrib. Mineral. Petrol. 144, 523–547.

Johnson, J.E., Gerpheide, A., Lamb, M.P., Fischer, W.W., 2014. O2 constraints from Paleoproterozoic detrital pyrite and uraninite. Geol. Soc. Am. Bull. 126, 813–830.

Kamber, B.S., Webb, G.E., Gallagher, M., 2014. The rare earth element signal in Archean microbial carbonate: information on ocean redox and biogenicity. J. Geol. Soc. 171, 745–763.

Kamo, S.L., Davis, D.W., 1994. Reassessment of Archean crustal development in the Barberton Mountain Land, South Africa. Tectonics 13, 167–192.

Kasting, J.F., 1992. Models relating to Proterozoic atmospheric and oceanic chemistry. In: Schopf, J.W., Klein, C. (Eds.), The Proterozoic Biosphere, A Multi-disciplinary Study. Cambridge University Press, Cambridge, UK, pp. 1185–1187.

Klinkhammer, G.P., Palmer, M.R., 1991. Uranium in the oceans: where it goes and why. Geochim. Cosmochim. Acta 55, 1799–1806.

Langmuir, D., 1978. Uranium solution–mineral equilibra at low temperatures with applications to sedimentary ore deposits. Geochim. Cosmochim. Acta 42, 547–569.

Langmuir, D., Herman, J.S., 1980. The mobility of thorium in natural waters at low temperatures. Geochim. Cosmochim. Acta 44, 1753–1766.

Li, W., Johnson, C.M., Beard, B.L., 2012. U–Th–Pb isotope data indicate Phanerozoic age for oxidation of the 3.4 Ga Apex Basalt. Earth Planet. Sci. Lett. 319–320, 197–206.

Li, W., Czaja, A.D., Van Kranendonk, M.J., Beard, B.L., Roden, E.E., Johnson, C.M., 2013a. An anoxic, Fe(II)-rich, U-poor ocean 3.46 billion year ago. Geochim. Cosmochim. Acta 120, 65–79.

Li, W., Huberty, J.M., Beard, B.L., Kita, N.T., Valley, J.W., Johnson, C.M., 2013b. Con-trasting behavior of oxygen and iron isotopes in banded iron formation revealed by in situ isotopic analysis. Earth Planet. Sci. Lett. 384, 132–143.

Lowe, D.R., 2013. Crustal fracturing and chert dike formation triggered by large me-teorite impacts, ca. 3.260 Ga, Barberton greenstone belt, South Africa. Geol. Soc. Am. Bull. 125, 894–912.

Lyons, T.W., Reinhard, C.T., Planavsky, N.J., 2014. The rise of oxygen in the Earth’s early atmosphere. Nature 506, 307–315.

Massey, M.S., Lezama-Pacheco, J.S., Jones, M.E., Ilton, E.S., Cerrato, J.M., Bargar, J.R., Fendorf, S., 2014. Competing retention pathways of uranium upon reaction with Fe(II). Geochim. Cosmochim. Acta 142, 166–185.

Olson, S.L., Kump, L.R., Kasting, J.F., 2013. Quantifying the areal extent and dissolved oxygen concentrations of Archean oxygen oases. Chem. Geol. 362, 35–43.

Partin, C.A., Lalonde, S.V., Planavsky, N.J., Bekker, A., Rouxel, O.J., Lyons, T.W., Kon-hauser, K.O., 2013. Uranium in iron formations and the rise of atmospheric oxygen. Chem. Geol. 362, 82–90.

Planavsky, N., Rouxel, O.J., Bekker, A., Hofmann, A., Little, C.T.S., Lyons, T.W., 2012. Iron isotope composition of some Archean and Proterozoic iron formations. Geochim. Cosmochim. Acta 80, 158–169.

Planavsky, N.J., Asael, D., Hofmann, A., Reinhard, C.T., Lalonde, S.V., Knudsen, A., Wang, X., Ossa Ossa, F., Recoits, E., Smith, A.J.B., Beukes, N.J., Bekker, A., Johnson, T.M., Konhauser, K.O., Lyons, T.W., Rouxel, O.J., 2014. Evidence for oxygenic pho-tosynthesis half a billion years before the Great Oxidation Event. Nat. Geosci. 7, 283–286.

Riding, R., Fralick, P., Liang, L., 2014. Identification of an Archean marine oxygen oasis. Precambrian Res. 251, 232–237.

Rouxel, O.J., Bekker, A., Edwards, K.J., 2005. Iron isotope constraints on the Archean and Paleoproterozoic ocean redox state. Science 307, 1088–1091.

Simonson, B.M., 2003. Origin and evolution of large Precambrian iron formation. In: Chan, M.A., Archer, A.W. (Eds.). Extreme deposition environments: mega end members in geologic time, vol. 370. Geological Society of America Special Paper, pp. 231–244.

Skulan, J.L., Beard, B.L., Johnson, C.M., 2002. Kinetic and equilibrium Fe isotope frac-tionation between aqueous Fe(III) and hematite. Geochim. Cosmochim. Acta 66, 2995–3015.

Stacey, J.S., Kramers, J.D., 1975. Approximation of terrestrial lead isotope evolution by a two-stage model. Earth Planet. Sci. Lett. 26, 207–221.

Stefurak, E.J.T., Lowe, D.R., Zenter, D., Fischer, W.W., 2014. Primary silica granules – a new mode of Paleoarchean sedimentation. Geology 42, 283–286.

Stookey, L.L., 1970. Ferrozine – a new spectrophotometric reagent for iron. Anal. Chem. 42, 779–781.

Tice, M.M., Bostick, B.C., Lowe, D.R., 2004. Thermal history of the 3.5–3.2 Ga On-verwacht and Fig Tree Groups, Barberton greenstone belt, South Africa, in-ferred by Raman microspectroscopy of carbonaceous material. Geology 32, 37–40.

Wazne, M., Korfiatis, G.P., Meng, X., 2003. Carbonate effects on hexavalent uranium adsorption by iron oxyhydroxide. Environ. Sci. Technol. 37, 3619–3624.

Weist, D., Wasserburg, G.J., 1987. Rb–Sr and Sm–Nd isotope geochemistry and chronology of cherts from the Onverwacht Group (3.5 AE), South Africa. Geochim. Cosmochim. Acta 51, 973–984.

Widdel, F., Schnell, S., Heising, S., Ehrenreich, A., Assmus, B., Schink, B., 1993. Ferrous iron oxidation by anoxygenic phototrophic bacteria. Nature 362, 834–836.

A.M. Satkoski et al. / Earth and Planetary Science Letters 430 (2015) 43–53 53

Wu, L., Percak-Dennett, E.M., Beard, B.L., Roden, E.E., Johnson, C.M., 2012. Stable iron isotope fractionation between aqueous Fe(II) and model Archean ocean Fe–Si coprecipitates and implications for iron isotope variations in the ancient rock record. Geochim. Cosmochim. Acta 84, 14–28.

Yamaguchi, K.E., Johnson, C.M., Beard, B.L., Ohmoto, H., 2005. Biogeochemical cy-cling of iron in the Archean–Paleoproterozoic Earth: constraints from iron isotope variations in sedimentary rocks from the Kaapvaal and Pilbara Cratons. Chem. Geol. 218, 135–169.