Research ArticleGEOCHEMISTRY

Sedimentary sulfur isotopes and Neoarchean ocean oxygenation

See allHide authors and affiliations

Science Advances  24 Jan 2018:
Vol. 4, no. 1, e1701835
DOI: 10.1126/sciadv.1701835


Abrupt disappearance of mass-independent fractionation of sulfur isotopes (MIF-S) from the geologic record and an apparent ingrowth in seawater sulfate around 2.45 billion years ago (Ga) signal the first large-scale oxygenation of the atmosphere [the Great Oxygenation Event (GOE)]. Pre-GOE O2 production is evident from multiple other terrestrial and marine proxies, but oceanic O2 concentrations remain poorly constrained. Furthermore, current interpretations of S isotope records do not explain a concurrent expansion in the range of both MIF-S—diagnostic for low atmospheric O2—and δ34S beginning at 2.7 Ga. To address these unknowns, we developed a reaction-transport model to analyze the preservation patterns of sulfur isotopes in Archean sedimentary pyrites, one of the most robust and widely distributed proxies for early Earth biogeochemistry. Our modeling, paradoxically, reveals that micromolar levels of O2 in seawater enhance the preservation of large MIF-S signals, whereas concomitant ingrowth of sulfate expands the ranges in pyrite δ34S. The 2.7- to 2.45-Ga expansion in both Δ33S and δ34S ranges thus argues for a widespread and protracted oxygenation of seawater, at least in shallow marine environments. At the micromolar levels predicted, the surface oceans would support a strong flux of O2 to the atmosphere, where O2 sinks balanced these fluxes until the GOE. This microoxic seawater would have provided habitat for early aerobic microorganisms and supported a diversity of new O2-driven biogeochemical cycles in the Neoarchean.


Two key features of sulfur isotope distributions in Archean pyrites underpin assertions for very low Archean atmospheric O2 concentrations: records of large mass-independent fractionations of sulfur isotopes (Δ33S; MIF-S) (13) and a narrow range of δ34S values (4). MIFs are produced in an ozone-free atmosphere, and atmospheric photochemical models suggest delivery of MIF-S to the Earth’s surface only when atmospheric O2 concentrations are less than 10−5 present atmospheric levels (1, 2). The relatively restricted ranges of δ34S in Archean pyrites, particularly before 2.7 billion years ago (Ga), point to low sulfate concentrations in the oceans, reflecting both restricted oxidative weathering of pyrites on land and little oxidative recycling of sulfur in the oceans (5, 6). Thus, the broad-scale features of the Δ33S and δ34S records paint a coherent picture of Archean atmospheric chemistry—low O2 concentrations that lead to both extensive photochemical sulfur cycling imparting large MIF and limited biogeochemical sulfur cycling at the Earth’s surface imparting small mass-dependent fractionations (MDFs). There is, however, a more nuanced structure to the Archean S isotope record including an expansion in the ranges of both Δ33S and δ34S between 2.7 and 2.45 Ga (Fig. 1). The Δ33S expansion was attributed to changes in the oxidation state of volcanic sulfur vented to the atmosphere (7), whereas the δ34S expansion was thought to reflect increased sulfate concentrations in oceans (4). Here, we show that the Neoarchean records of Δ33S and δ34S both indicate progressive seawater oxygenation beginning at 2.7 Ga, and our models constrain the corresponding levels of O2 and sulfate in Archean coastal environments.

Fig. 1 Changes in Δ33SFeS2, εSO4-FeS2, oceanic sulfate, and oxygen throughout the Archean.

Record of (A) Δ33SSO4-FeS2 and (B) εSO4-FeS2 (assuming a seawater sulfate value of 12‰) and the simulated (C) oceanic sulfate and (D) O2 levels through time. Data points are from Canfield and Farquhar (5) and Johnston (33). Simulated outlines for sulfate and O2 concentrations correspond to values for which our model reproduces the S isotope records. My, million years.


The δ34S and Δ33SFeS2 isotopic records preserved in Archean sedimentary pyrites carry signatures of atmospheric processes as well as information about isotopic alterations taking place in oceanic water column and in sediments after their deposition. The transformations during early diagenesis were investigated here with a one-dimensional reaction-transport model, which traced the differential dynamics of sulfur isotopes through a network of geochemical and biogeochemical reactions (see Methods and the Supplementary Materials) and provided insights into the preservation patterns of S isotopes and the role of oxygen in them.

The pre-GOE (Great Oxygenation Event) expansion in sulfate-pyrite δ34S isotopic differences, εSO4-FeS2, has been previously ascribed to rising seawater sulfate concentrations, which lead to more intense sedimentary sulfate reduction and generate isotopically lighter pyrite (4, 8). Our simulations support this conclusion but additionally suggest that the presence of dissolved O2 also increases the δ34S range by intensifying in-sediment sulfate-sulfide redox cycling. Low-micromolar O2 concentrations, however, only increase εSO4-FeS2 by a few per mil, because sulfur cycling remains limited and fractionation during sulfide oxidation is minor in comparison to that imparted during sulfate reduction (Fig. 2A). The effect of O2 on the preservation of Δ33S signatures, in contrast, is pronounced. Oxygen in the atmosphere is generally thought to preclude the formation of MIF-S due to ozone formation and the shielding of the ultraviolet radiation needed for SO2 photolysis (9, 10). Our modeling, on the other hand, reveals that O2 in the ocean enhances the preservation of MIF-S in pyrite (Fig. 2B). Pyrite is typically formed when iron monosulfide reacts with either elemental S or hydrogen sulfide (RFeS+S and RFeS+HS in table S1). When the atmospheric MIF-S signal is transmitted from elemental S to pyrite through diagenesis, it is thus influenced by the isotopic composition of iron monosulfide and porewater hydrogen sulfide (fig. S1). Because hydrogen sulfide can be produced from non–MIF-S sulfate through microbially catalyzed reactions such as sulfate reduction that impart MDF, its addition to pyrite dilutes the MIF-S signal. Hydrogen sulfide precipitation with ferrous iron generates iron monosulfide, the isotopic composition of which becomes a complex function of the intensity of sulfate reduction, recycling of hydrogen sulfide through redox processes within sediment, and Rayleigh distillation (11, 12). These processes make the MIF-S signatures in pyrite dependent on the rate of sulfate reduction, which is regulated by the availability of sulfate and reactive organic matter (13), and several other factors involved in sediment S cycling (Figs. 2B and 3 and fig. S6). Oxygen plays a key role by controlling oxidative S cycling and the supply of reactive organic matter available to fuel sulfate reduction. In particular, higher O2 levels enhance the preservation of MIF-S signals (Fig. 2B) by decreasing the sulfate reduction rates through deepening of the depth of the sulfate reduction zone in the sediment and limiting the supply of reactive organic matter reaching that zone. In contrast to the effect of O2, increased sulfate levels result in more intense sulfate reduction and hydrogen sulfide production, and more intense dilution of the MIF-S signal, resulting in lower Δ33S (Figs. 2B and 3 and figs. S6 and S7).

Fig. 2 Modeled range of εSO4-FeS2 and Δ33SFeS2 for different oxygen and sulfate concentrations.

(A) Modeled range of δ34S isotopic differences between seawater sulfate and sedimentary pyrites (εSO4-FeS2) for different O2 and sulfate levels. (B) Modeled range of Δ33SFeS2 for which the atmospheric MIF-S signal is preserved in sedimentary pyrites, as a function of seawater O2 and sulfate levels. Up-down arrows correspond to the respective ranges of MIF and MDF. Higher-resolution details at the low sulfate range are provided in fig. S6.

Fig. 3 Effect of oxygen, sulfate, organic matter, imposed Δ33SO4, and Δ33S0 on εSO4-FeS2 and Δ33SFeS2.

(A) Sulfur isotope data (circles) in Archean pyrites older than 2.4 Ga (33). Sulfate-pyrite isotopic differences εSO4-FeS2 were calculated from rock δ34S values for the value of δ34S in seawater sulfate of 12‰. Shaded region corresponds to the model’s range in Fig. 2, obtained for sediment organic carbon (OC) content of 0.1%, the Δ33S0 in the deposited elemental sulfur of 10‰, and the Δ33SO4 in seawater sulfate of zero. Arrows indicate the directions in which the modeled range shifts in response to increases in indicated parameters. (B and C) Ranges (dark shading) of O2 and sulfate that correspond to the isotopic ranges within the respective outlines I and II in (A). Lightly shaded areas correspond to the respective ranges in individual isotopic parameters, εSO4-FeS2 and Δ33SFeS2, calculated for the same set of parameters as in Fig. 2. The modeled trends indicate that the preservation of large εSO4-FeS2 values observed after 2.7 Ga (Fig. 1) requires high sulfate, whereas the preservation of large Δ33SFeS2 values requires the presence of O2.

Isotopic variability induced by rising marine O2 and concomitant increases in seawater sulfate provide a self-consistent explanation for the observed Neoarchean expansions in Δ33S and δ34S ranges. In Archean rocks, high values of Δ33S tend to be associated with low values of εSO4-FeS2, whereas larger values of εSO4-FeS2 tend to occur at low Δ33S (Fig. 3). This pattern has been attributed to variability in the atmospheric processes that produce MIF-S (7). Figure 3 illustrates that diagenetic processes can impart a similar relationship, where δ34S and Δ33S covary depending on seawater O2 and sulfate concentrations. The increase in Δ33S variability in the Neoarchean was previously attributed to changes in the oxidation state of volcanic sulfur emitted to the atmosphere (7) and to the associated shift in photochemical S transformation pathways (7, 14, 15). This explanation, however, relies on the assumption that atmospherically produced MIF-S signals remain unmodified by diagenesis (7), which requires persistently negligible sediment sulfate reduction. This conflicts with recent evidence for increased microbial sulfate reduction in the Neoarchean (16) and with the correlative expansion in εSO4-FeS2, which is a known indicator for increased sulfate concentrations and enhanced rates of sulfate reduction (4). Furthermore, there remains no known geological process that would induce the inferred shift in the oxidation state of volcanic gases before the GOE. Our model suggests instead that the Neoarchean expansion in Δ33S and δ34S reflects the protracted oxygenation of seawater and an early ingrowth in marine sulfate (Fig. 3).

Comparisons of our model results with the δ34S and Δ33S compositions of Archean pyrites allow us to constrain the O2 and sulfate levels of contemporaneous seawater and identify the effects of several diagenetic factors on the preserved sulfur isotopic signatures (Fig. 3). The concentrations of O2 and sulfate required to reproduce given the δ34S and Δ33S ranges depend on factors such as sediment organic matter concentration and the isotopic values of elemental S and sulfate delivered to the sediment (Fig. 3). The narrow mid-Archean ranges of both δ34S and Δ33S require pyrite formation under very low O2 and low sulfate concentrations. The preservation of large values of Δ33S observed after 2.7 Ga, in contrast, requires a rise in seawater O2 concentrations to several micromolar. The correlative increase in εSO4-FeS2 signals a contemporaneous increase in seawater sulfate to concentrations on the order of 200 μM. Increased sulfate alone, however, does not reproduce the εSO4-FeS2 record, which also requires enhanced deposition of OC (Fig. 3) because, even at low sulfate levels, sulfate reduction becomes co-limited by organic substrates (13). The Neoarchean S isotope expansion, therefore, likely signals not only the onset of marine oxidative sulfur cycling but also an increase in the ocean’s biological productivity, at least in localized coastal oxygen oases.

S isotopes thus paint a cogent picture of Neoarchean ocean oxygenation that is supported by insight from other proxies. Abundant evidence implies O2 production across more than 700 million years before the GOE (6, 1721). Development of ocean O2 “oases” at the onset of the Neoarchean era with concentrations up to tens of micromolar is supported by inferences from the geochemistries of molybdenum (18) and manganese and iron (22, 23), whereas cerium anomalies in carbonates (23) imply shallow water O2 concentrations approaching ~5 μM O2. Our model supports these estimates and shows that micromolar O2 concentrations in the ocean are compatible with (2427) and are possibly even required by the MIF-S record. Global biogeochemical models reveal that seawater O2 concentrations on the order of 1 to 10 μM (27), and possibly up to 25 μM, can exist under an ostensibly anoxic atmosphere, despite large fluxes of O2 from the ocean (28). Further, the Neoarchean development of coastal euxina (6) may also reflect an accompanying increase in seawater sulfate concentrations resulting from enhanced oxidative continental weathering. Our results further suggest that broadening of the δ34S range during the GOE, after the disappearance of the MIF-S signal (~2.5 to 2.4 Ga; Fig. 1), reflects an increase in seawater O2 from low micromolar concentrations to beyond a threshold value of tens of micromolar. The sensitivity of pyritic εSO4-FeS2 values to O2 is dependent on microbial physiology but increases markedly in the range of 10 to 15 μM (see the Supplementary Materials). The approximately 10‰ increase in εSO4-FeS2 observed at the GOE thus likely indicates a transition from low micromolar seawater O2 concentrations to several tens of micromolar, rather than its first production and accumulation.

The Neoarchean era was thus characterized by a major reorganization of the global S cycle that saw an ingrowth of seawater sulfate, the onset of oxidative sediment sulfur cycling, and an increase in the role of sediment sulfate reduction beginning around 2.7 Ga. This reorganization, captured in the isotopic composition of carbonate-associated sulfate (16), made microbial sulfate reduction an important sink for marine sulfate (16). The onset of O2-driven marine sulfur cycling, as revealed by the sulfur isotope record (5), signals a broader expansion of the oxic marine biosphere and the proliferation of aerobic microbial metabolisms likely linked to heterotrophic carbon respiration and possibly methanotrophy and oxidative nitrogen cycling in the sea.


Geochemical and isotopic modeling

Transformations of S from seawater sulfate and MIF-carrying elemental S deposited from the atmosphere were simulated at steady state with a diagenetic equationEmbedded Image(1)

The reaction network Rij for chemical species Ci considered sulfate reduction, reoxidation of the produced hydrogen sulfide, precipitation of iron monosulfide and subsequent formation of pyrite, and, variably, disproportionation of elemental S (fig. S1). Vertical transport in the sediment matrix of porosity φ was characterized by molecular diffusion coefficients Di and burial velocity v neglected for solutes. A complete description of the model is given in the Supplementary Materials; the reactions and their rate formulations are listed in tables S1 and S2, and model parameter values are listed in table S3. The rate of sulfate reduction, a microbially mediated reaction known to impart the largest S isotopic fractionations, was simulated using Michaelis-Menten kinetics (table S2). The corresponding half-saturation concentration of sulfate, Km, was on the order of low micromolar, consistent with values in modern low-sulfate environments (13), and the rate expression included a term that described inhibition of sulfate reduction in the presence of O2. Isotopic fractionations during diagenetic reactions were calculated using specified fractionation factors (fig. S1). Patterns of the S isotope preservation were calculated for seawater sulfate concentrations between 1 and 400 μM and under a range of seawater O2 concentrations between 0 and 30 μM. The effects of varied seawater O2 and sulfate concentrations on pyrite δ34S were expressed as deviations (εSO4-FeS2) relative to the value in sulfate from overlying water, which was set to 12‰ as the benchmark value for Archean seawater (5); alternative values are considered in the Supplementary Materials. To investigate the behavior of atmospherically derived MIF-S signals, our model assumed a flux of elemental S with a specified Δ33S composition and predicted the fate of this MIF-S signal in sedimentary pyrite. Because the aerosol elemental S Δ33S values were not likely to have been spatially uniform and the corresponding values of δ33S and δ34S in elemental S are not well constrained for the Archean, we set the isotopic composition of S0 to Δ33SS0 = 10‰ as a benchmark value (see Supplementary Materials) and varied this in a sensitivity analysis. Seawater sulfate was assumed to carry no MIF signal (Δ33SSO4 = 0) in benchmark simulations, but the model was also run for an alternative set of values (Δ33SSO4 = −7‰) (9). In comparison to previous models for the Archean S cycle (4, 8), the present model expanded the range of modeled processes by including an oxidative cycle and elemental S disproportionation and by explicitly considering the kinetics of pyrite precipitation. Although timing the evolutionary appearance of microbial disproportionation is debated (29), we included it in the model to test for its possibility (4, 30, 31) and then verified that its absence does not modify our main conclusions. Similar to previous models (4, 8), our model did not consider sulfate reduction coupled to anaerobic methane oxidation because its contribution to sulfate reduction in the Neoarchean was likely low given the slow kinetics of anaerobic methane oxidation at low sulfate concentrations, as observed in modern low-sulfate lakes (8, 32). Furthermore, the corresponding S isotopic fractionations during methane oxidation are poorly known, confounding their description in our model (32).


Supplementary material for this article is available at

section S1. Geochemical model

section S2. Isotopic model

section S3. Sensitivity analysis

table S1. Reactions included in the model.

table S2. Kinetics of the reactions included in the model.

table S3. Parameters for reactive transport modeling and sensitivity analysis.

table S4. Thermodynamic favorability of reactions suggested for elemental sulfur formation.

table S5. Model boundary conditions for sulfate, sulfide, and oxygen.

fig. S1. Sediment sulfur cycle with isotopic fractionation associated with each reaction.

fig. S2. Typical sediment profiles of sulfate, sulfide, oxygen, iron, CRS pool including FeS and elemental sulfur, and rates of reduction (RSR) and reoxidation (RSOX) along with isotopic fractionation of sulfate, sulfide, and pyrite for Archean conditions.

fig. S3. Change in εSO4-FeS (‰) in response to changes in model parameter values.

fig. S4. Sensitivity of model outputs to organic matter availability at [O2] = 1 μM.

fig. S5. Sensitivity of model outputs to organic matter availability at [SO4] = 10 μM.

fig. S6. Effect of oxygen inhibition constant (Ki) on εSO4-FeS at [SO4] = 10 μM.

fig. S7. Modeled range of Δ33SFeS2 for which the atmospheric MIF-S signal is preserved in sedimentary pyrites, as a function of seawater oxygen and sulfate levels.

fig. S8. Effect of sulfate concentration on Δ33SFeS2 at [O2] = 10 μM.

fig. S9. Effect of organic matter availability on Δ33SFeS2 at [SO4] = 10 μM and [O2] = 10 μM.

fig. S10. Effect of oceanic oxygen level on Δ33SFeS2 at [SO4] = 10 μM.

References (3458)

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank C. Reinhard and an anonymous reviewer for comments that helped improve the manuscript. Funding: This work was partially supported by Natural Sciences and Engineering Research Council of Canada Discovery grant 0487 to S.A.C. Author contributions: M.F. performed modeling, interpreted results, and contributed to writing the manuscript as a first author. S.A.C. contributed ideas, helped place results in the context of available data, and contributed to writing. S.K. supervised the project, interpreted results, and contributed to writing. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.
View Abstract

Navigate This Article