Abstract

The accumulation of oxygen in Earth’s atmosphere and oceans in the late Archean had profound implications for the planet’s biogeochemical evolution. Oxygen impacts sulfur cycling through the oxidation of sulfide minerals and the production of sulfate for microbial sulfate reduction (MSR). The isotopic signature of sulfur species preserved in the geologic record is affected by the prevailing biological and chemical processes and can therefore be used to constrain past oxygen and sulfate concentrations. Here, in a study of a late Archean analogue, we find that the sulfur isotopic signature in the water column of a seasonally stratified lake in southern China is influenced by MSR, whereas model results indicate that the isotopic signature of the underlying sediments can be best explained by concurrent sulfate reduction and sulfide oxidation. These data demonstrate that small apparent sulfur isotope fractionations (δ34Ssulfate-AVS = 4.2‰–1.5‰; AVS—acid volatile sulfides) can be caused by dynamic sulfur cycling at millimolar sulfate concentrations. This is in contrast to current interpretations of the isotopic record and indicates that small fractionations do not necessarily indicate very low sulfate or oxygen.

INTRODUCTION

Evidence for oxygen production in the oceans dates from 2.8 b.y. ago (Farquhar et al., 2011), although oxygen did not significantly accumulate in the atmosphere until the Great Oxidation Event (GOE) 2.4 b.y. ago (Farquhar and Wing, 2003; Bekker et al., 2004). This temporal offset raises questions about the importance of oxidation reactions with the reduced iron and sulfur present in the Archean oceans. One way of constraining oxygen levels is through the stable isotope composition of sulfur preserved in the geologic record, as increasing atmospheric oxygen concentrations led to an increase in oxidative weathering processes that enhanced the delivery of sulfate to the oceans and allowed a more dynamic, microbially driven sulfur cycle. This in turn affected the isotopic composition of sulfur because microbial sulfate reduction (MSR) and disproportionation of partially oxidized sulfur compounds (S0, S2O32−, or SO32−) have a strong impact on the isotopic composition of the reduced and oxidized sulfur pools (Johnston et al., 2005a; Canfield and Thamdrup, 1994). The isotopic composition of sulfur preserved as pyrite in the rock record compared to sulfate (as evaporites or trace sulfate in carbonates), Δ34S, has therefore been used extensively to reconstruct the emergence and scope of these processes throughout geologic time (Habicht et al., 1998, 2002; Shen et al., 2003; Crowe et al., 2014): 
graphic

Archean sedimentary rocks generally have lower Δ34S values than sedimentary rocks from later periods in Earth’s history (cf. Havig et al., 2017). One explanation for this observation is that seawater sulfate concentrations (and atmospheric O2) during the Archean were low. At low sulfate concentrations, the magnitude of the apparent fractionation imparted during MSR may be suppressed due to physiological effects (Habicht et al., 2002; Bradley et al., 2016) or due to reservoir effects caused by sulfate depletion (Crowe et al., 2014; Gomes and Hurtgen, 2013, 2015). Moreover, low fractionation may indicate minimal sulfide oxidation because disproportionation of intermediate sulfur species formed during sulfide oxidation typically enhances the apparent isotope fractionation. This approach assumes that the isotopic signature formed in the water column is preserved in the sediment, and has led to estimates of Archean seawater sulfate concentrations of a few to hundreds of micromolar (Crowe et al., 2014; Habicht et al., 2002).

In this study, we demonstrate that the isotopic signature formed in the water column of a seasonally stratified lake (Aha Reservoir in southern China) with sulfate concentrations up to 1.5 mM is not preserved in the sediments. Contrary to the current paradigm, a reoxidative sulfur cycle in the sediment lessens the apparent fractionation between sulfide and sulfate that is expressed in the water column, despite the absence of reservoir effects.

METHODOLOGY

All samples were preserved using standard procedures (see references below, and further details in the GSA Data Repository1). The concentrations of sulfur species [ΣS(-II), S0, S2O32−, SO32−, SO42−] were determined using established analytical procedures (see the Data Repository). Multiple-sulfur-isotope composition was measured after Ono et al. (2006) on samples precipitated as Ag2S. Isotopic composition is presented in per mil using standard δ notation relative to Vienna Canyon Diablo troilite (VCDT): 
graphic
in which 3xR = 3xS/32S (x = 3 or 4).
The minor isotope composition is presented using Δ33S notation, which describes the deviation of δ33S from a reference fractionation line (Farquhar et al., 2003): 
graphic

Detailed descriptions of the sampling, sample preservation, and concentration and isotopic analyses, as well as of model design, are provided in the Data Repository.

BIOGEOCHEMICAL CYCLING IN AHA RESERVOIR

Aha Reservoir is a seasonally stratified lake located in southern China (106°37′W, 26°34′N; Item DR1 in the Data Repository; Fig. 1). The 24-m-deep water column was stratified at the time of sampling (10–12 August 2016), with dissolved O2 depleted by 6 m depth and sulfide undetectable to 22 m depth (Fig. 2A). Total Mn concentrations were as much as 22 µM, total Fe concentrations were <2 µM, and SO42− concentrations were relatively constant with depth (Fig. 2A). Sulfide concentrations increased to 80 µM above the sediment-water interface. The sediments consisted of brown to black mud, with high total and reactive Fe content (Fig. DR2) and relatively high total organic carbon content (2.5%–4.5%).

The water column contained high concentrations of Mn relative to Fe (Fig. 2A), whereas the sediments had higher Fe than Mn content (Fig. DR2). This division is driven by the different reactivity of Fe and Mn toward sulfide. Both Mn and Fe oxides react with sulfide to form divalent metal ions and S0: 
graphic
 
graphic
yet there are two key differences that led to the preferential retention of Fe in the sediment. First, the reaction between sulfide and MnO2 is approximately four times faster than the reaction with Fe(O)OH (Yao and Millero, 1993, 1996). Second, Fe2+ interacts differently with sulfide than reduced Mn. MnS is rarely formed, thus Mn2+ diffuses from the sediment into the water column. In contrast, FeS forms readily at circum-neutral pH, thereby trapping Fe in the solid phase within the sediments and leading to the observed geochemical differences between the water column and sediments (Fig. 2; Fig. DR2). The effect of these different geochemical settings is that sulfur cycling and isotope dynamics in the water column and sediments are controlled by different processes.

DECOUPLING BETWEEN SULFUR ISOTOPE COMPOSITION IN THE WATER COLUMN AND IN THE SEDIMENT

The isotopic composition of sulfate above 7 m depth was similar to the isotopic composition of sulfate in rainwater (Fig. 2B; δ34Srainwater = –7‰; Song et al., 2011). The constant concentrations and δ34S value of –10.5‰ ± 2.1‰ between 7 and 20 m water depth are consistent with coal-derived sulfate, the main source of sulfate to the lake (δ34Scoal = –12‰; Song et al., 2011). An increase in δ34Ssulfate and the presence of sulfide below 22 m depth indicates that microbial sulfate reduction was active within the water column. This activity results in δ34Ssulfate – δ34Ssulfide34Ssulfate-sulfide) from 17‰ to 30‰ and Δ33Ssulfate-sulfide of –0.04‰ to –0.09‰, within the range of fractionations produced by pure cultures of sulfate-reducing microorganisms.

Sulfate concentrations in the surface sediment pore water (0–4 cm, Fig. 2C) were higher than in the lake water column, despite constant chloride concentrations that are consistent with water column values (Fig. DR3). This indicates that sulfide oxidation in the upper sediment exceeds sulfate reduction and that non-steady-state conditions with high seasonal variability persist in the surface sediments. We suggest that Fe(III) (hydr)oxides are formed in the surface sediment during the winter, when the water column is mixed and oxygenated. In the early summer, these high concentrations of reactive iron (hydr)oxides result in reoxidation of sulfide produced by MSR. During the late summer and autumn, sulfate reduction likely prevails over sulfide oxidation. The decrease in sulfate concentrations below 6 cm combined with the high concentrations of reduced sulfur species (Fig. 2C) indicates that MSR is the net process at these depths. SO42− concentrations increased notably below 20 cm, concurrent with decreasing amounts of S0, acid volatile sulfides (AVS), and chromium reducible sulfur (CRS), suggesting that extensive oxidation of sulfide also occurs deeper in the sediment.

Sulfide oxidation is reflected in the stable sulfur isotope composition in the sediment (Figs. 2D and 3). In the surface sediment, the observed fractionation between SO42− and reduced sulfur falls within the range for MSR; however, the relationship between α34 and α33 (i.e. the fractionation factor for 34S and 33S associated with a specific process) is inconsistent with MSR (δ34Ssulfate-AVS [i.e., δ34Ssulfate – δ34SAVS] = 1.5‰–4.2‰, Δ33Ssulfate-AVS = 0.03‰–0.05‰), suggesting that the sulfur cycle is not driven solely by sulfate reduction. Furthermore, SO42− in the surface sediment (δ34Ssulfate = –13.7‰ to –18.4‰) was isotopically lighter in 32S than both bulk water column SO42−34Ssulfate = –10.5‰) and water column SO42− affected by MSR (δ34Ssulfate = 0.58‰). Even more intriguing are the positive Δ33Ssulfate-AVS values in the surface sediments (Δ33Ssulfate-AVS = 0.03‰–0.05‰; Fig. DR4), although models and observations corroborate a negative value for this parameter during MSR (Farquhar et al., 2007; Zerkle et al., 2010; Sim et al., 2011; Johnston et al., 2005b). We are only aware of two previous reports of positive Δ33Ssulfate-AVS values. First, positive Δ33Ssulfate-AVS values were reported during sulfur isotope fractionation by disproportionating microorganisms at higher δ34Ssulfate-AVS values than observed here (Johnston et al., 2005b). Second, the combination of positive Δ33Ssulfate-AVS values and low positive δ34Ssulfate-AVS values was attributed to fast, abiotic sulfur cycling in a hydrothermal pool (Cinder Pool, Yellowstone National Park, Wyoming, USA; Kamyshny et al., 2014).

To explore the possible combination of processes leading to the observed fractionations in the sediments of Aha Reservoir, we used a simple box model based on the metabolic model of Farquhar et al. (2007) that considers the theoretical isotopic fractionation (both δ34S and Δ33S) during different combinations of sulfate reduction (associated with high fractionation), intermediate disproportionation (associated with high fractionation), and sulfide oxidation (associated with low fractionation) in the upper 4 cm of the sediment. These processes connect three different sulfur pools: AVS (total labile sulfide), Si (an unspecified sulfur intermediate), and SO42−. Oxidation of AVS forms Si and SO42−. Sulfate reduction is modeled as a one-step reaction that results in formation of AVS (Fig. DR5).

We applied this model to three sets of biogeochemical processes: (1) AVS oxidation (without sulfate reduction or disproportionation), (2) AVS oxidation followed by disproportionation of Si, and (3) sulfate reduction and AVS oxidation (Fig. 4). The first two scenarios do not agree with the observed isotopic values (Fig. 4; Item DR3); however, sulfide oxidation combined with sulfate reduction (>50‰) without contribution from disproportionation fully encompass the observed values. δ34Ssulfate-AVS and Δ33Ssulfate-AVS in the surface sediment of Aha Reservoir are therefore consistent with a scenario in which sulfide oxidation plays an important role, along with sulfate reduction. The sulfur cycle in the sediment thus creates an isotopic signature distinct from that formed in the water column.

IMPLICATIONS FOR THE INTERPRETATION OF STABLE SULFUR ISOTOPES

In contrast to previous studies, we found that an oxidative sulfur cycle in the sediments of Aha Reservoir reduces the observed difference between the isotopic composition of reduced sulfur in the sediment and both pore water and surface water SO42−. The prevailing paradigm regarding sulfide oxidation in sediment is that sulfide is chemically oxidized by Fe(III) (hydr)oxides to S0, which is then microbially disproportionated, resulting in a characteristically large isotopic difference between sulfide and sulfate (Canfield and Thamdrup, 1994; Böttcher et al., 2001). However, although S0 disproportionation driven by Fe(III) (hydr)oxides can result in a large apparent isotopic fractionation between sulfide and sulfate, the degree of this apparent fractionation decreases if sulfide oxidation to SO42− occurs, for example driven microbially or by MnO2 (Böttcher and Thamdrup, 2001). Our results clearly indicate that a broader view of potential sulfide oxidation processes and their isotopic signatures throughout geologic time is necessary.

Paleo-oceanographic interpretations rely upon Δ34S (Equation 1), which compares the isotopic composition of sedimentary sulfide with SO42− from surface waters. In modern low-sulfate systems, Δ34S is typically lower than the instantaneous microbial fractionation (εSR) due to reservoir effects caused by Rayleigh distillation of a small sulfate pool. The value Δ34S thus appears to be dependent upon SO42− concentrations between 0 and 5 mM (Gomes and Hurtgen, 2015), with smaller Δ34S values observed at lower SO42− concentrations.

The SO42− concentrations in Aha Reservoir were 1.5 mM in the water column; therefore we should be able to calculate the SO42− concentration in Aha Reservoir using Δ34S. Pyrite in the deeper sediment of Aha Reservoir has an isotopic value of δ34SCRS = –21.9‰ and the average bulk water column sulfate is –10.5‰, yielding a Δ34S of 11.4‰. Based upon this value, the model of Gomes and Hurtgen (2015) calculates SO42− concentrations of 290 µM in Aha Reservoir, which is lower than the actual concentrations (∼1.5 mM). This is because the system is not at steady state and the isotope fractionation in Aha Reservoir cannot be explained using Rayleigh models. Extensive reoxidative sulfur cycling in the sediment affects the value of Δ34S so that it is no longer diagnostic for the water column concentration of SO42−. Although sulfide oxidation results in low Δ34S values, similar to those resulting from reservoir effects (Crowe et al., 2014), the underlying biogeochemical processes are distinct and reflect different environmental conditions with respect to sulfate concentration and oxidant availability.

The results presented here provide important insight into the processes affecting the preservation of isotopic signatures in sediments, including the role of sulfide oxidation. The sedimentary Fe concentrations in Aha Reservoir are high compared to modern marine sediments, but the processes described above are common to all environments in which sulfide is produced in the presence of oxidized Fe or Mn minerals. The removal of large quantities of Fe from ancient ferruginous oceans due to widespread oxidation and the subsequent formation of iron deposits during the late Archean and early Proterozoic may have created an environment similar to that of Aha Reservoir, with sediments highly enriched in Fe(III) (hydr)oxides (Bekker et al., 2014). The low bulk Δ34S values observed during this time (<10‰; Havig et al., 2017) may therefore be compatible with higher sulfate concentrations and more extensive sulfide oxidation than previously considered.

ACKNOWLEDGMENTS

We acknowledge support from a joint Israeli Science Foundation (ISF)–National Natural Science Foundation of China (NSFC) grant to Kamyshny (ISF 2214/115) and Guo (NSFC 41561144005 and 41625006); a Fulbright Postdoctoral Fellowship, a Kreitman Fellowship, and a Marie Curie European Fellowship (SedSulphOx 746872) to Findlay; and funding from the Danish National Research Foundation (DNRF104) and the European Research Council (294200) (Pellerin). We thank Guangxu Zhu, Zhongyi Zhang, and Jingjing Zhao for help with fieldwork, Nir Ben Eliahu and Hadar Cohen for laboratory assistance, and James Farquhar, William Leavitt, and Bo Barker Jørgensen for comments on the manuscript. We thank the anonymous reviewers of this manuscript for their helpful comments.

1GSA Data Repository item 2019266, methods and model description and equations, is available online at http://www.geosociety.org/datarepository/2019/, or on request from editing@geosociety.org.
Gold Open Access: This paper is published under the terms of the CC-BY license.