Transient surface ocean oxygenation recorded in the ∼2.66-Ga Jeerinah Formation, Australia

Significance

Understanding how and when Earth’s surface became oxygenated is essential for understanding its biogeochemical evolution. Incipient oxygenation of Earth’s surface environments before the Great Oxidation Event (GOE; ∼2.4 Ga) has been well-documented, but the nature of these redox changes, whether protracted or transient, is poorly understood. We present nitrogen isotope ratios, selenium abundances, and selenium isotope ratios from the Jeerinah Formation (∼2.66 Ga; Fortescue Group, Western Australia) that represent (i) high-resolution evidence of transient surface ocean oxygenation ∼260 My before the GOE, (ii) a possible muted pulse of oxidative continental weathering, and (iii) the oldest firm evidence for nitrification and denitrification metabolisms. These results, in concert with previous studies, highlight the variability in mechanisms and magnitudes of Neoarchean oxygen fluctuations.

Abstract

Many paleoredox proxies indicate low-level and dynamic incipient oxygenation of Earth’s surface environments during the Neoarchean (2.8–2.5 Ga) before the Great Oxidation Event (GOE) at ∼2.4 Ga. The mode, tempo, and scale of these redox changes are poorly understood, because data from various locations and ages suggest both protracted and transient oxygenation. Here, we present bulk rock and kerogen-bound nitrogen isotope ratios as well as bulk rock selenium abundances and isotope ratios from drill cores sampled at high stratigraphic resolution through the Jeerinah Formation (∼2.66 Ga; Fortescue Group, Western Australia) to test for changes in the redox state of the surface environment. We find that both shallow and deep depositional facies in the Jeerinah Formation display episodes of positive primary δ¹⁵N values ranging from +4 to +6‰, recording aerobic nitrogen cycling that requires free O₂ in the upper water column. Moderate selenium enrichments up to 5.4 ppm in the near-shore core may indicate coincident oxidative weathering of sulfide minerals on land, although not to the extent seen in the younger Mt. McRae Shale that records a well-documented “whiff” of atmospheric oxygen at 2.5 Ga. Unlike the Mt. McRae Shale, Jeerinah selenium isotopes do not show a significant excursion concurrent with the positive δ¹⁵N values. Our data are thus most parsimoniously interpreted as evidence for transient surface ocean oxygenation lasting less than 50 My, extending over hundreds of kilometers, and occurring well before the GOE. The nitrogen isotope data clearly record nitrification and denitrification, providing the oldest firm evidence for these microbial metabolisms.

Despite widespread agreement about general trends (1), our understanding of the mode and tempo of Earth’s oxygenation still suffers from considerable uncertainties. Among the more salient features of this record is a proposed “whiff” of oxygen in surface environments at 2.5 Ga, well before the onset of the Great Oxidation Event (GOE) at ∼2.3–2.4 Ga (2, 3), which has been identified using combined trace metal and nitrogen isotope datasets collected at high stratigraphic resolution (4–6). Exploration of other Meso- and Neoarchean records has revealed dynamic evolution of surface ocean redox conditions and oxidative continental weathering leading up to the permanent establishment of oxidizing conditions at Earth’s surface (7–21). To further elucidate the dynamic redox landscape in the Neoarchean, we conducted a coupled survey of nitrogen and selenium isotopes and abundances at high stratigraphic resolution in two drill cores spanning a near-shore–offshore transect across the ∼2.66-Ga Jeerinah Formation in Western Australia. These complementary proxies allow us to assess the relative magnitude and spatial extent of redox fluctuations.

Nitrogen is essential for life as we know it, and its biogeochemical behavior is highly redox sensitive. Since at least 3.2 Ga, it is evident that the biosphere has exerted the primary control on Earth’s surface nitrogen fluxes, regulating nitrogen availability and speciation in the ocean and atmosphere (22, 23). Metabolisms that drive nitrogen speciation in the ocean (transition of nitrogen among different redox states) can use a variety of electron donors for reduction, such as organic matter, Fe²⁺, and sulfide, but for oxidation, are primarily reliant on dissolved O₂ as an electron acceptor due to the high oxidation potential (Eh) of nitrate (24). Ammonium oxidation to nitrate can occur down to low nanomolar oxygen concentration (25, 26), which is a fraction of a percent of modern well-oxygenated ocean waters. Thus, a significant pool of nitrate in an ancient ocean would suggest that at least minor amounts of oxygen were freely available to allow for nitrification: the biologically mediated conversion of ammonium to nitrite/nitrate.

Modern microbial metabolisms are known to have a diverse range of effects on the relative abundances of the nitrogen isotopes ¹⁴N and ¹⁵N (summarized in table 1 of ref. 24; refs. 27, 28). For example, N₂ fixation using Mo-nitrogenase causes an insignificant isotopic fractionation during the conversion of N₂ to organic-bound nitrogen, but biological denitrification of NO₃⁻ to N₂ strongly prefers ¹⁴N, producing isotopically light N₂ gas and an isotopically heavy residual dissolved NO₃⁻ pool. Hence, nitrogen isotopes in ancient sedimentary rocks have allowed paleoecological interpretations of nitrogen metabolisms in ancient ecosystems. These interpretations are only semiquantitative due to overlapping fractionation factors for different metabolic pathways and because of the potential for isotopic resetting by postdepositional alteration. Still, the presence of nitrate in concentrations high enough for significant bioassimilation can be distinguished isotopically from anaerobic settings that are dominated by nitrogen fixation followed by proximal reassimilation of liberated ammonium (5, 22–24, 29–31).

We also studied the behavior of selenium (Se), a redox-sensitive element with oxyanions that are stable at similar redox potentials to nitrogen oxyanions, to determine whether changes in marine nitrogen cycling were concurrent with changes in weathering dynamics on land. Unlike nitrate, which is primarily produced during biomass remineralization and nitrification in the water column, Se oxyanions are mainly sourced from oxidative weathering on land (32). Additionally, if Se fluxes are sufficiently large and oxygenated waters are sufficiently widespread, Se isotopes can become fractionated during incomplete reduction of Se oxyanions (33). The nitrogen and selenium proxies combined can thus inform us about the loci of oxygenation.

Geologic Setting

As part of the Agouron Institute Drilling Project (AIDP) in 2012, two ultraclean diamond drill cores AIDP-2 and AIDP-3 were sampled across the boundary between the Hamersley and Fortescue Groups in Western Australia. These holes were collared at 21°16′51″ S, 120°50′02″ E (AIDP-2) and 21°46′32″ S, 117°34′11″ E (AIDP-3) across a basinal depth gradient, with AIDP-2 intersecting shallower near-shore facies and AIDP-3 intersecting deeper offshore facies. AIDP-2 consists of ∼158 m of the Carawine Dolomite that, for this study, was sampled at low stratigraphic resolution (SI Appendix, Fig. S1) and ∼128 m of predominantly black shales from the Jeerinah Formation that were sampled at 1-m resolution. AIDP-3 is ∼138-m deep and capped by the basal Marra Mamba Iron Formation, but samples were taken exclusively from the underlying Jeerinah Formation at 1-m resolution. The shallower depositional environment of AIDP-2 is indicated by the stromatolitic carbonates of the Carawine Dolomite (34) and by the shoreface setting of the basal Woodiana Member of the Jeerinah Formation, which consists of occasionally stromatolitic orthoquartzites (35). In contrast, AIDP-3 transects only a basinal banded iron formation and sulfidic kerogenous shales. The two holes can be stratigraphically correlated using a horizon of meteorite impact spherules (36), which occur in a ∼20-m-thick megabreccia toward the bottom of the Carawine Dolomite in AIDP-2 and in a 1-cm-thick graded spherule layer in the upper Jeerinah Formation in AIDP-3 (SI Appendix, Fig. S2). Rocks from both AIDP-2 and AIDP-3 are generally thought to be within the prehnite–pumpellyite metamorphic facies (37), but an alternative upper greenschist facies interpretation exists (38). However, this latter interpretation is inconsistent with the pelitic mineralogy of the AIDP shales, because key index minerals of upper greenschist conditions (e.g., biotite, chloritoid) are absent. Also, it depends on a chlorite geothermometer (39) that (i) unjustifiably assumes that all iron in the chlorite is Fe²⁺, which can lead to temperature overestimations (40), and (ii) yields infeasibly broad temperature ranges (>140 °C) for samples in close stratigraphic proximity. Hence, we favor the prehnite–pumpellyite metamorphic facies assignment (addressed further in SI Appendix).

Results

Se abundances are tightly scattered around an average of 1.2 ± 1.1 ppm throughout both cores, but a moderate excursion up to 5.4 ppm occurs around 363 m in AIDP-2. After normalizing the data to Al, total carbon percentage (TOC), and total sulfur (TS), the Se excursion in AIDP-2 still appears statistically significant (Fig. 1 B–D and SI Appendix, Figs. S3 and S4). Multiple regression confirms this inference, showing that less than one-half of the observed Se enrichment can be explained by lithological variability (R² = 0.49) (Fig. 1 B–D). δ^82/78Se shows no systematic trend across either of the two cores (average +0.3 ± 0.3‰). In contrast, δ¹⁵N displays a large range from −2 to +14‰, with systematic shifts detected by our linear regression model. This identified three distinct stages (from I to III from oldest to youngest) in each core (Table 1). The only exception to the model boundary placement is the boundary between stages I and II in AIDP-3, where the boundary position was handpicked based on marked changes in TOC, total nitrogen percentage (TN), and carbon/nitrogen (C/N) ratios (atomic). Importantly, this boundary still falls within the range of stratigraphic positions where the model predicts a significant change in δ¹⁵N_bulk values.

Fig. 1.

Cross-plots of (A) δ¹⁵N_kerogen values and C/N ratios, (B) Se enrichment factor (EF) and Se (parts per million), (C) Se (parts per million) and TOC, and (D) Se (parts per million) and TS (percentage). For A, note the linear covariance in AIDP-3 stage I compared with the rest of the core. Samples from AIDP-2 come from all stages. For B, the procedure for calculating EF is described in SI Appendix. Note the particularly strong enrichments in AIDP-2 stage II. Samples from AIDP-3 come from all stages. For C and D, note the weaker correlations in AIDP-2 compared with AIDP-3. The variation (and therefore, the increase) in selenium abundance in AIDP-2 cannot be explained through variations in TOC and TS. Conversely, much of the variation in selenium abundance in AIDP-3 can be explained by variations in TOC and TS. This suggests that the increase in Se during stage II of AIDP-2 is likely caused by an increased flux of Se to the basin.

Table 1.

Stratigraphic stages of AIDP-2 and AIDP-3 with associated nitrogen and carbon measurements

Stage	Strat., m	δ15Nbulk, ‰	δ13Corg, ‰	TN, %	TOC, %	C/N*	TIC, %
AIDP-2
III	344.18–278.75	+0.38 ± 0.61	−43.2 ± 2.5	0.08 ± 0.03	4.8 ± 2.3	81 ± 39	11.4 ± 4.4
II	377.89–344.99	+4.33 ± 0.53	−43.0 ± 1.9	0.12 ± 0.02	7.7 ± 2.4	74 ± 17	9.5 ± 2.5
I	428.16–380.37	+2.76 ± 0.71	−42.0 ± 1.5	0.04 ± 0.01	2.8 ± 0.8	95 ± 24	16.3 ± 4.9
AIDP-3
III	127.83–66.03	+2.39 ± 0.86	−41.7 ± 1.5	0.06 ± 0.02	5.8 ± 2.4	109 ± 32	12.1 ± 8.6
II	155.75–129.65	+6.07 ± 0.36	−41.7 ± 1.0	0.06 ± 0.01	4.7 ± 0.7	85 ± 14	15.0 ± 3.0
I	204.37–156.89	+10.2 ± 2.14	−44.2 ± 0.9	0.02 ± 0.02	2.4 ± 2.1	156 ± 63	14.7 ± 2.9

Stratigraphic-stage boundaries were determined using the linear regression model described in the SI Appendix. Strat., stratigraphy; TIC, acid-soluble mass-fraction estimate of total inorganic carbon.

^*C/N ratios are given as atomic ratios.

AIDP-2.

Stage I in AIDP-2 is characterized by the most variable δ¹⁵N_bulk values in the core, averaging around +2.8‰. In this stage, TOC and TN are at their lowest. From stage I to stage II, δ¹⁵N_bulk suddenly increases by an average of ∼1.5‰ to the highest values recorded in the core (+5.43‰) and becomes less variable. This rise in δ¹⁵N_bulk values is associated with increases in both TOC and TN by about a factor of three. From stage II to stage III, δ¹⁵N_bulk values decrease by 4‰ on average to a minimum of −0.71‰. TOC and TN also decrease across this transition but not to levels as low as in stage I. δ¹⁵N_kerogen values are systematically lower than δ¹⁵N_bulk measurements throughout the core by an average of 1.6 ± 0.9‰, and therefore, they show the same general trend as δ¹⁵N_bulk values across all three stages (Fig. 2).

Fig. 2.

Chemostratigraphy of AIDP-2 showing δ¹⁵N_bulk (gray dots), δ¹⁵N_kerogen (red dots), C/N ratios, δ^82/78Se, and Se concentrations. Stages are delineated by horizontal dotted lines. The shaded area for δ^82/78Se represents crustal values. Down-core depth is given in meters. BIF, banded iron formation.

AIDP-3.

Stage I in AIDP-3 has the most positive and variable δ¹⁵N_bulk values and C/N (atomic) ratios in either core, averaging +11.0‰ and 174, respectively. Notably, in this same stage, TOC and TN averages are the lowest in either core. Across the boundary between stages I and II, δ¹⁵N_bulk values decrease by 5‰ on average (Fig. 3). TOC and TN values both increase by over a factor of three, and C/N (atomic) ratios decrease to an average similar to all stages apart from stage I of AIDP-3. The transition from stage II to stage III is defined by another decrease in δ¹⁵N_bulk of about 4‰ on average, similar to the stage II to stage III transition in AIDP-2. TOC, TN, and C/N (atomic) ratios do not change significantly from stage II to stage III.

Fig. 3.

Chemostratigraphy of AIDP-3 showing δ¹⁵N_bulk (gray dots), δ¹⁵N_kerogen (red dots), C/N ratios, δ^82/78Se, and Se concentrations. Stages are delineated by horizontal dotted lines. Lithological symbols are the same as in Fig. 2. Note the different δ¹⁵N and C/N scales between AIDP-2 and AIDP-3, which are required to accommodate stage I of AIDP-3. The shaded area for δ^82/78Se represents crustal values. Down-core depth is given in meters.

Like AIDP-2, in AIDP-3, δ¹⁵N_kerogen values are systematically lower than δ¹⁵N_bulk. Whereas the average difference between the two measurements is consistent throughout AIDP-2, in AIDP-3, it varies in a stepwise fashion throughout the core; the smallest differences are in stage III, and the largest differences are in stage I. A noteworthy relationship in stage I is that δ¹⁵N_kerogen values correlate strongly with atomic C/N_kerogen and C/N_bulk ratios (Fig. 1A). This is not the case in stages II and III and all of AIDP-2.

Discussion

Proxy Alteration.

Postdepositional processes, such as diagenesis and metamorphism, have the potential to alter the primary nitrogen and Se isotopic signals preserved in sedimentary rocks. For example, Se isotope fractionation could potentially occur during diagenetic oxyanion reduction. However, this would require the presence of a stable Se oxyanion reservoir in the overlying water column, which would in turn imply oxic conditions. Such diagenetic Se reactions would, therefore, strengthen environmental redox signals. During metamorphism, Se is fairly immobile (41), and therefore its isotopic ratios are unlikely to be altered. Nitrogen isotopes, however, can undergo more significant alteration during diagenesis and metamorphism (42). The δ¹⁵N of sinking and sediment biomass can increase during early diagenesis under oxic conditions through the preferential remineralization, oxidation, and loss of ¹⁴N (common in modern marine settings), but it is less susceptible to alteration under the mainly anoxic bottom waters/sediments (43–45) characteristic of the Neoarchean. It is possible that, for samples with initially positive δ¹⁵N_bulk values, early diagenesis under anoxic conditions can decrease δ¹⁵N_bulk values by the addition of isotopically light (δ¹⁵N of −2 to +1‰) biomass via the in situ growth of nitrogen-fixing bacteria; samples that have a primary δ¹⁵N_bulk signal around 0‰ would not be significantly affected by this mechanism (46–48). Primary δ¹⁵N values may also decrease by the preferential deamination of ¹⁵N-enriched organic compounds, such as proteins (46). Even if more relevant in the Neoarchean, neither of these mechanisms can fully account for the shifts in δ¹⁵N_bulk values of up to 4‰ between stages in both cores, and they would not affect our interpretations regarding episodes of enriched δ¹⁵N values.

Anoxic diagenesis may lead to small isotopic differences between the kerogen and silicate nitrogen fractions and even smaller isotopic changes in bulk rock measurements, but metamorphism/metasomatism can lead to greater isotopic disparity and changes that are not always predictable (49, 50). Below greenschist facies, such thermal effects can lead to changes in the primary nitrogen isotopic signal of bulk rocks by 1–2‰ (42). As previously mentioned, the Carawine Dolomite and Jeerinah Formation were likely metamorphosed to prehnite–pumpellyite grade (37) and therefore, may have been subject to these small changes to the primary δ¹⁵N values. As metamorphic grade and metasomatic alteration increase, so does the difference in δ¹⁵N between bulk rock and kerogen fractions, which may reach up to 13‰ (51).

To test if metamorphism/metasomatism has affected our samples, we considered the relationship between δ¹⁵N values and C/N ratios as well as the difference in δ¹⁵N values between the two nitrogen fractions (kerogen and silicate). If thermal processes have significantly affected our samples, increasing δ¹⁵N values should correspond to increasing C/N ratios, as nitrogen is more mobile than carbon and ¹⁴N is more mobile than ¹⁵N under such conditions. The only samples that show this relationship come from stage I in AIDP-3 (Fig. 1A). This is also the stage where the differences in δ¹⁵N values between the two nitrogen fractions are the greatest and most variable. Because of this, stage I of AIDP-3 evidently displays a secondary isotopic signal due to thermal alteration. This most likely had a metasomatic component, as the kerogen samples are not persistently more altered (higher C/N and δ¹⁵N) with stratigraphic depth but show local variability that could have resulted from focused fluid–rock interactions. This is consistent with previous observations that underlying basalt units locally experienced metasomatic alteration (52). Indeed, the basalt underlying the Jeerinah Formation in AIDP-3 has metamorphic/metasomatic textures, including zones of K-feldspar spherulites; veins and amygdales filled with chlorite, calcite, and silica; and patches of sericitization, all indicating multiple periods of metasomatic alteration (SI Appendix, Fig. S5). Because all samples from AIDP-2 and from stages II and III of AIDP-3 lack petrographic and geochemical signs of significant thermal alteration, the δ¹⁵N values from these stages are interpreted to be close to primary compositions.

Interpretation of Nitrogen Isotopic Data.

δ¹⁵N values between −2 and +1‰, as seen in AIDP-2 stage III, are generally interpreted to reflect marine nitrogen cycling dominated by nitrogen fixation using Mo-nitrogenase (53). Elevated δ¹⁵N values greater than +1‰, as in all other stages, can be interpreted in one of three ways (30, 31) (SI Appendix, Fig. S6). (i) Partial bioassimilation from a dissolved NH₄⁺ pool that would preferentially remove ¹⁴N locally creates isotopically light biomass and an isotopically heavy residual NH₄⁺ pool. If this heavy residual pool was transported to another location in the basin, it could be fully assimilated, and biomass there would record isotopically heavy δ¹⁵N values (54). However, because AIDP-2 (shallow) and AIDP-3 (deep) represent a broad cross-basinal transect and because no corresponding isotopically light values (less than −2‰) were encountered anywhere, there is no evidence to suggest that this process was responsible for the heavy δ¹⁵N values in either core. (ii) Partial nitrification of an ammonium pool that preferentially selected ¹⁴N followed by quantitative denitrification of the resulting nitrate would leave the residual ammonium isotopically heavy. Bioassimilation from this ammonium pool would result in isotopically heavy biomass (55). (iii) Quantitative nitrification of an ammonium pool followed by partial denitrification of the resulting nitrate pool that preferentially removed ¹⁴N would leave residual nitrate isotopically enriched in ¹⁵N (5, 29). Assimilation of this heavy nitrate pool would yield δ¹⁵N-enriched biomass. This mechanism is largely responsible for the enriched δ¹⁵N values of nitrate in the modern deep ocean that are spatially heterogeneous (ranging from +2 to +20‰) but show an average around +5‰ (56). The lack of very light δ¹⁵N values suggests that mechanism ii or iii, both requiring free O₂ for nitrification, best explains our data. Mechanism ii should yield highly variable δ¹⁵N values, because it lacks an isotopically heavy nitrogen sink from the ocean (31), but such variability is not seen in most of the AIDP stages. We thus propose that mechanism iii probably prevailed. In either case, the isotopically enriched nitrogen data require oxic conditions in marine surface waters.

Interpretation of Selenium Concentrations and Isotopic Ratios.

The extent of oxygenation can be further elucidated using Se isotopes and abundances. The peak in Se abundance during stage II of the AIDP-2 core may reflect (i) enhanced oxidative terrestrial weathering or (ii) enhanced delivery of volcanically derived SeO₂. We cannot definitively rule out a volcanic Se source for the Se enrichment in AIDP-2, as the volcanic flux of Se to marine environments may have been comparable with or greater than the Se flux from oxidative weathering under low O₂ conditions (32). We would, however, expect that such an event should have been more uniformly distributed across the basin and recorded in AIDP-3. As AIDP-3 does not show any significant Se enrichment, our data are perhaps more consistent with a small fluvial Se influx that was sequestered near shore under dominantly anoxic bottom waters. In any case, the <6-ppm Se enrichment is noticeably smaller than that during the 2.50-Ga pulse of oxygen recorded in the Mt. McRae Shale (6), where Se abundance reaches 20 ppm. This difference suggests that the Se flux to the ocean was considerably less during the Jeerinah oxygenation event. This conclusion is further supported by the Jeerinah Se isotopic data, which nearly all fall within the crustal range (from −0.3 to +0.6‰) (57), indicating that any Se oxyanion reservoir was small and nearly quantitatively sequestered in near-shore environments; this is unlike the Mt. McRae Shale, which shows a marked positive excursion in δ^82/78Se to +1.4‰ (6). Therefore, while the nitrogen data are comparable between the two events, the Se data suggest that the magnitude of oxidative weathering was smaller at 2.66 Ga than at 2.50 Ga.

Transient Surface Ocean Oxygenation in the Neoarchean: Mode and Tempo.

A transient enrichment in Archean sedimentary δ¹⁵N at 2.5 Ga has previously been linked to an oxygen whiff before the GOE (5), and this AIDP dataset extends such evidence for surface ocean oxygenation back by ∼160 Ma. Godfrey and Falkowski (29) previously reported several elevated δ¹⁵N measurements from a roughly coeval section in the Transvaal Supergroup in South Africa sampled at coarse stratigraphic resolution, which are also consistent with aerobic nitrogen cycling in the surface ocean. However, the δ¹⁵N data presented here show that surface ocean oxygenation was a transient phenomenon lasting 50 My or less (see calculations below). It also provides direct evidence that this episode of oxygenation extended over hundreds of kilometers, with surface waters evidently holding nanomolar to micromolar concentrations of dissolved O₂, which allowed widespread nitrification (25, 26) as previously predicted (58).

The mode of oxygenation cannot be resolved using nitrogen isotopes alone. Aerobic nitrogen cycling requires bioavailable trace metals, such as Mo, Cu, and Fe, along with free O₂ and dissolved NH₄⁺, and it is not clear which was limiting before and after stage II. Surface ocean oxygenation could have been a purely oceanographic event driven by changes in upwelling that delivered nutrients like Fe and NH₄⁺ to the photic zone, stimulating cyanobacterial productivity (58). Alternatively, mild oxidative weathering on land during stage II could have supplied bioessential trace metals to basin surface waters without supplying much Se due to its high Eh. This could have resulted from localized oxidative weathering beneath nonmarine cyanobacterial mats, hypothesized to have been significant before the GOE (59), or from a small global pulse in atmospheric oxygen levels. If cyanobacteria and nitrifiers were trace metal limited (60, 61), either mechanism would spur regional oxygen production, allowing free O₂ and NO₃⁻ to accumulate in the photic zone. Both are consistent with the positive δ¹⁵N values in stage II of the two cores as well as the observed Se enrichment in AIDP-2.

If the Se enrichment in AIDP-2 was due to weak oxidative weathering of continental material (our preferred interpretation), then oxygenation of the surface ocean and mobilization of Se oxyanions were out of phase as indicated by the δ¹⁵N and [Se] values. Regardless of whether surface ocean oxygenation resulted from delivery of nutrients by upwelling or oxidative weathering, the Se peak could indicate a period of more vigorous oxidative continental weathering, but not to the extent of the 2.5-Ga Mt. McRae whiff. The occurrence of similarly positive nitrogen isotope values in rocks of similar age from South Africa indicates that surface oceanic oxygenation leading to aerobic nitrogen cycling (29) may have been relatively widespread at this time. Fluctuating balances between atmospheric O₂ sources and sinks could have led to transient increases in atmospheric O₂ of various magnitudes, consistent with a more global positive δ¹⁵N signal and the Se enrichment. Alternatively, terrestrial cyanobacterial mat proliferation under a transient organic haze (18, 62, 63) providing temporary UV protection (64) may have increased the oxidative weathering flux of Se, but the feedback between O₂ generation, aerobic methane oxidation, and haze formation needs to be explored in greater detail.

As mass independent fractionation of sulfur isotopes is persistent throughout the Jeerinah Formation (65, 66), any atmospheric oxygenation event could have only been of limited magnitude. The duration of the δ¹⁵N excursion in AIDP-2, covering ∼35 m of shale deposition, suggests that the shifting balance between O₂ sources and sinks driving oxygenation would have operated on a geologically rapid (<50-My) timescale. [The top and bottom of the Jeerinah Formation are dated to 2.629 ± 5 (67) and 2.684 ± 6 (68), respectively (differences of 44 Ma minimum and 66 Ma maximum). The positive δ¹⁵N values in stage II occur in ∼30 m of black shale in the middle of the Jeerinah in both cores, accounting for <25% of the formation’s total thickness. It is, therefore, likely that oxygenation occurred on a timescale perhaps much less than 50 Ma, assuming similar deposition rates between the two sections.] As the Se abundance excursion is even more stratigraphically restricted (<20 m), any interval of atmospheric oxygenation would have been proportionately shorter.

The δ¹⁵N data also clearly indicate that the microbial metabolisms of nitrification and denitrification had already evolved by ∼2.66 Ga (SI Appendix, Fig. S7), confirming the few data points previously reported from similarly ancient rocks from South Africa (29). Both nitrification and denitrification are complex pathways involving many redox steps utilizing multiple metalloenzymes, which on the modern Earth, are typically mediated by microbial consortia or in rare cases, by individual microbial taxa (69, 70). This implies that, by the Mid-Neoarchean, microbes had developed either evolutionary or ecological sophistication comparable with their modern counterparts.

Conclusion

The data presented here record a second Archean event characterized by transient oxygenation of the surface ocean. While the nitrogen isotopic signal is similar in magnitude to the 2.50-Ga Mt. McRae whiff of oxygen, the Jeerinah event shows a smaller selenium enrichment and no isotopic excursion in δ^82/78Se, suggesting that terrestrial oxidative weathering was weaker during this earlier episode. When viewed at high resolution, these data show that oxygenated surface waters both preceded and postdated the oxidative weathering signal of selenium, perhaps indicating that local production of oxygen eventually enabled more vigorous sulfide dissolution on land. These results highlight the variability in mechanisms and magnitudes of Neoarchean oxygen fluctuations and show the importance of feedbacks between biogenic oxygen production, oxidative weathering, and nutrient fluxes during the prelude to the permanent oxygenation of Earth’s surface environment.

Methods

Sample preparation and analytical procedures for bulk rock analyses of nitrogen isotopes followed the methodology outlined in refs. 31 and 71, and Se abundances and isotopes were determined using the protocol described in ref. 72; details are in SI Appendix. The linear regression model used to identify the stratigraphic location of significant shifts in δ¹⁵N_bulk followed the methodology reported in ref. 73 and is described in greater detail in SI Appendix.