A productivity collapse to end Earth’s Great Oxidation

Significance

The Great Oxidation Event (GOE) ca. 2,400 to 2,050 Ma caused the first significant accumulation of free oxygen in the atmosphere and potentially a dramatic growth of oxidant reservoirs on the Earth’s surface in a suggested “oxygen overshoot.” However, the termination of this event remains poorly understood. Here, we present geochemical data suggesting a drastic decline in gross primary productivity across the end-GOE transition, delineating a shift from “feast” to “famine” conditions characteristic of the next 1 billion y.

Abstract

It has been hypothesized that the overall size of—or efficiency of carbon export from—the biosphere decreased at the end of the Great Oxidation Event (GOE) (ca. 2,400 to 2,050 Ma). However, the timing, tempo, and trigger for this decrease remain poorly constrained. Here we test this hypothesis by studying the isotope geochemistry of sulfate minerals from the Belcher Group, in subarctic Canada. Using insights from sulfur and barium isotope measurements, combined with radiometric ages from bracketing strata, we infer that the sulfate minerals studied here record ambient sulfate in the immediate aftermath of the GOE (ca. 2,018 Ma). These sulfate minerals captured negative triple-oxygen isotope anomalies as low as ∼ −0.8‰. Such negative values occurring shortly after the GOE require a rapid reduction in primary productivity of >80%, although even larger reductions are plausible. Given that these data imply a collapse in primary productivity rather than export efficiency, the trigger for this shift in the Earth system must reflect a change in the availability of nutrients, such as phosphorus. Cumulatively, these data highlight that Earth’s GOE is a tale of feast and famine: A geologically unprecedented reduction in the size of the biosphere occurred across the end-GOE transition.

The rise of oxygen in Earth’s atmosphere during the early Paleoproterozoic was one of the most transformative events in all of Earth’s history. Evidence for this event can be observed through the disappearance of mass-independently fractionated sulfur isotopes (1) within reduced and oxidized forms of sulfur ca. 2,430 to 2,330 Ma (2) as well as macroscale features in the sedimentary record such as the emergence of red beds or the disappearance of detrital pyrite and uraninite (3). Following the initial rise in atmospheric oxygen is the largest positive shift in the carbonate carbon isotope record [∼2,220 to 2,060 Ma (4, 5)] termed the Lomagundi–Jatuli Excursion (LJE). The interval between the initial rise of O₂ ca. 2,430 to 2,330 Ma and the end of the LJE—marked by carbon isotope values returning to values of ∼0‰—has traditionally defined Earth’s Great Oxidation Event (6) and we follow this convention here.

The LJE has widely been interpreted as a transient rise in organic carbon burial and by consequence a rise in atmospheric O₂. Importantly, this rise in O₂ has been suggested to exceed not only Archean, but also background Proterozoic (6, 7) and possibly Phanerozoic levels in a so-called “oxygen overshoot” (8, 9). However, this high-O₂ interpretation can be tempered under different assumptions regarding changes in the isotopic value of carbon inputs to the global dissolved inorganic carbon (DIC) reservoir as well as the fractionation associated with carbon fixation by primary producers (10). Moreover, some researchers question whether these ca. 2,220- to 2,060-Ma carbonates actually record changes to global marine DIC and instead have suggested that such isotopic records may document local and/or diagenetic processes (e.g., ref. 11). This controversy surrounding the LJE has motivated independent tests of whether such an oxygen overshoot, and corresponding transition from a high-pO₂ syn-Great Oxidation Event (GOE) state to a comparatively low-pO₂ post-GOE state, even occurred. Numerous independent proxies, while differing in degree of severity, broadly characterize this interval as a decline in surface environment oxidant inventories (12–18). However, it is worth noting that many such records may better reflect local depositional environments and not necessarily global conditions. Moreover, almost all of these records reflect the passive response of a geochemical proxy to a change in Earth’s oxidant reservoir (i.e., the effect), rather than capturing the atmospheric signal of interest or, perhaps more importantly, the underlying cause. Cumulatively, it is clear that the application of an independent test of such models would prove useful in shedding light on the end-GOE transition.

A critical factor in evaluating the GOE and the end-GOE transition is an understanding of the mechanism driving oxygenation. A rise in atmospheric oxygen during the GOE could have been the result of increased organic carbon production (i.e., increased gross primary production [GPP] and by consequence a larger biosphere). To sustain a larger biosphere during both the buildup to and maintenance of a high-pO₂ state during the GOE, it has been proposed that high weathering rates (perhaps including siderite; ref. 9) and consequent phosphorus release would have resulted from elevated H₂SO₄ generation via pyrite oxidation during the GOE (although this remains debated; ref. 12). This large release of phosphorus would have then sustained the posited oxygen overshoot in a feast-like scenario (6, 8). In marked contrast to the elevated GPP hypothesized for the GOE, evidence has been presented that characterizes the mid-Proterozoic [∼2,000 to 1,100 Ma (14)] as an interval of remarkably low GPP, possibly just 6% of modern levels (14, 19). Although GPP changes across this transition have been assumed (16), quantifying the degree to which the GOE deviated from the mid-Proterozoic GPP state remains largely unexplored, and how quickly such a transition in the biosphere occurred to mark the end of the GOE remains unknown. To better understand the end-GOE transition and test the hypothesis that the end-GOE was brought about by a nutrient famine, we employ a combined isotope approach utilizing triple-oxygen (Δ¹⁷O), multiple-sulfur (δ³⁴S, Δ³³S, Δ³⁶S), and Ba isotopes (δ^138/134Ba) on barites collected from the Belcher Group, Nunavut, Canada. These samples capture the interval of time immediately after the GOE (Orosirian; ref. 20) and allow quantitative constraints to be placed on the productivity (GPP) of the biosphere at this time.

Geological Context.

The Belcher Group is a 7- to 10-km thick sedimentary basin in subarctic Canada, largely deposited between 2,018.5 ± 1.0 Ma and 1,854.2 ± 1.6 Ma (Fig. 1 and ref. 20). The lowermost Belcher Group is composed of ∼1 km of dolomicrite deposited in a sabkha environment (Kasegalik Formation). This is followed by several hundred meters of Eskimo Formation basalts. Overlying this are 3 km of generally supratidal to shallow subtidal carbonate sedimentary rocks with minor siliciclastic rocks (Fairweather, McLeary, Tukarak, Mavor, Costello, and Laddie formations). Progradation is recorded in the Rowatt and Mukpollo formations, followed by the deposition of granular iron formation in the Kipalu Formation. The overlying Flaherty Formation is composed of submarine basalt and is up to several kilometers thick. The Belcher Group concludes with the deposition of a flysch (Omarolluk Formation) and molasse [Loaf Formation (20–22)].

Fig. 1.

Location of the Belcher Islands, sampling site, and sulfate/evaporite occurrences. (A) Stratigraphic column of the Belcher Group, indicating the intervals at which gypsum pseudomorphs (gypsum pseudo.), halite casts, microbarite, and macrobarite were observed. Refer to SI Appendix, Fig. S1 for photographs of each type of occurrence. The Kasegalik Formation, deposited in a sabkha environment, was host to the largest amount of evaporite casts and SO₄ minerals observed in the Belcher Group. Higher in the stratigraphy, the McLeary Formation contains 1 stratigraphic horizon with a small amount of gypsum pseudomorphs. Finally, the Costello Formation contains centimeter-scale barite crystals along a single stratigraphic horizon on eastern Tukarak Island. (B) Map of North America with Belcher Islands in the black box. (C) Map of Belcher Islands. Macrobarites are indicated by a red circle. (D) Macrobarites, with Canadian penny for scale.

Evaporite casts and pseudomorphs of SO₄-bearing minerals have been reported in a number of stratigraphic intervals in the Belcher Group and have been further constrained during our fieldwork (Fig. 1 and refs. 22 and 23). In the Kasegalik Formation, chert-replaced gypsum rosettes and crystals are common in dolomicrites, with halite casts occurring more rarely in maroon, argillaceous dolomicrites. Standard heavy mineral separation techniques on 5- to 10-cm thick beds of sandstone from the Kasegalik Formation yielded a small fraction of microbarites (∼200 μm in width). No halite was observed in the McLeary Formation but it did bear small chert-replaced, twinned gypsum crystals along a single stratigraphic horizon <2 m thick. Finally, large (∼1–4 cm) barite crystals are present in the Costello Formation (22) and are the primary focus of this study. This formation is composed of some 10 to 20 m of gray to black shale, overlain by several hundred meters of rhythmically bedded, variably red, pink, gray, and cream-colored dolomicrite with shale partings and very thinly interbedded limestone. In a small area on eastern Tukarak Island, milky-white barite crystals up to several centimeters in width occur in massive, cream-colored dolomicrite (Fig. 1). There is no evidence for subaerial exposure or the presence of a strongly evaporitic/restricted environment associated with the Costello Formation barites; to the contrary, slumps and partial Bouma sequences have led to the interpretation that the Costello Formation was deposited in a foreslope environment (21).

Mass-Independent Oxygen-Isotope Variations.

The ∆¹⁷O anomaly recorded within sedimentary SO₄ reflects both the amount of tropospheric O₂ incorporated into SO₄ during pyrite oxidation [f_O2; ∼8% to 15% (24)] and the ∆¹⁷O value of ambient tropospheric O₂, which itself is driven by 3 principle variables: pCO₂, pO₂, and GPP (SI Appendix, Fig. S2). Photochemical reactions in the stratosphere involving the production of O₃ (ozone) do so with a mass-independent partitioning of isotopes, where O₃ becomes enriched in ¹⁷O and O₂ depleted in ¹⁷O, relative to a suggested definition of mass dependence [δ¹⁷O/δ¹⁸O = 0.5305 (25, 26)]. Depletions and enrichments in ¹⁷O relative to this definition of mass-dependent fractionation are termed “triple-oxygen isotope anomalies,” which are defined as ∆¹⁷O = δ¹⁷O − 0.5305(δ¹⁸O) and are reported on the permil (‰) scale relative to Vienna Standard Mean Ocean Water (V-SMOW). The ¹⁷O depletion in O₂ (negative ∆¹⁷O values) can be made larger via isotopic exchange between O₃ and CO₂ (27) where larger depletions are observed at higher pCO₂ levels. Mass-independently depleted O₂ produced in the stratosphere is then transported to the troposphere where it is mixed with photosynthetically derived O₂ from the biosphere that carries a mass-dependently fractionated ∆¹⁷O value [i.e., ∆¹⁷O of ∼0‰ (ref. 28 and SI Appendix, Fig. S2)]. Therefore, the degree of depletion from stratospheric reactions with CO₂, the rate of nonmass independent O₂ production from the biosphere, and the size of the O₂ reservoir where these fluxes compete set the ∆¹⁷O value of tropospheric O₂ (29). Provided that limited amounts of microbial cycling occur, tropospheric ∆¹⁷O values can be deposited and preserved in the sedimentary record in the form of SO₄-bearing minerals (e.g., gypsum, barite, CAS). All postatmospheric processes will shift values toward seawater (∆¹⁷O = 0‰), so ∆¹⁷O values within SO₄-bearing minerals that are more negative than modern tropospheric O₂ (∆¹⁷O = −0.516‰; ref. 30) must have been deposited under different pCO₂-pO₂-GPP conditions than those of the modern environment (14, 19, 29, 31, 32).

Sulfur-Isotope Variations.

The sulfur isotopic composition (δ³⁴S, Δ³³S) of SO₄-bearing minerals reflects several factors: The isotopic composition of sulfur inputs into the depositional system, which, on long- and short-term timescales is dominated by volcanically and sedimentary-derived sulfur, respectively; the type and intensity of microbial cycling that processes S in SO₄ to HS⁻ and eventually to pyrite through reactions with Fe in sediments; and any isotopic effects associated with precipitation itself (14, 33). Oxidative weathering of volcanically derived sulfur produces SO₄ with a δ³⁴S value (δ³⁴S = [^34/32R_sample − ^34/32R_standard]/^34/32R_standard) of ∼0‰ and a ∆³³S value (∆³³S = δ³³S − [δ³⁴S − 1]^0.515) of ∼ −0.05‰ (13) reported here relative to Vienna Cañon Diablo Troilite (V-CDT) scale. Fractionation by dissimilatory SO₄ reduction or microbial sulfide oxidation (34), which likely dominated microbial S cycling across the Paleoproterozoic (35), results in sulfide species with lower δ³⁴S values and thus leaves residual SO₄ isotopically enriched in ³⁴S (e.g., ref. 36). Minor sulfur isotopes are also sensitive to microbial cycling that will slightly increase or decrease ∆³³S values in residual SO₄, depending upon the dominant S metabolism (35–37). Such processes give the modern marine SO₄ reservoir a δ³⁴S value of +21.2‰ and a ∆³³S value of +0.05‰ (33). Therefore, multiple sulfur isotopes can aid in exploring the nature of ancient SO₄ deposits (e.g., marine or terrestrial) as well as provide insights into the degree of microbial-S cycling.

Barium-Isotope Variations.

Barium stable isotopes, δ^138/134Ba, can be used to trace the source and cycling of Ba to modern and ancient barite deposits (e.g., refs. 38 and 39; δ^138/134Ba = ^138/134Ba_sample/^138/134Ba_{NIST 3104a} − 1). Broadly speaking, barite formation involves the meeting of 2 segregated fluids at an interface (e.g., ref. 40) or in barite supersaturated microzones within otherwise undersaturated environments (e.g., ref. 41). Barite precipitation favors incorporation of isotopically light Ba with an isotopic effect ∼ −0.5‰ (e.g., ref. 41), rendering residual dissolved Ba enriched in “heavy” Ba isotopes by a corresponding amount. Thus, the isotopic composition of Ba in barites is sensitive to the size of the Ba reservoir. Barites produced from semi-infinite “open” reservoirs, such as seawater, exhibit a narrow range of compositions close to the crustal average composition (∼0.0‰ ± 0.1‰). In contrast, precipitates formed from finite or “closed” reservoirs, such as cold seeps or hot springs, can exhibit distinctive compositions (38). Such precipitates can exhibit δ^138/134Ba values with a non-0 mean and considerable isotopic variation, which is thought to arise from differences in the isotopic composition of the underlying local Ba source and its respective evolution under semirestricted settings, respectively. Thus, the mean and range of Ba isotope compositions of precipitated barite are broadly indicative of the source of Ba.

Materials and Methods

Nine “macrobarite” samples were collected from the Costello Formation for geochemical analyses. For ∆¹⁷O measurements, samples were dissolved and purified to produce pure barite, which was subsequently lazed and measured by isotope ratio mass spectrometry (IR-MS). For δ³⁴S and Δ³³S measurements, S was extracted and purified using Thode solution and then fluorinated and measured by IR-MS. Finally, δ^138/134Ba measurements were carried out by sample alkaline dissolution, barium purification using column chromatography, and subsequent analysis by multicollector inductively coupled plasma mass spectrometry. Refer to SI Appendix for detailed methods.

Results

Triple-oxygen- and Ba-isotope data for microbarites from the Kasegalik Formation and macrobarites from the Costello Formation along with multiple sulfur isotope data from the Costello Formation are summarized in SI Appendix, Table S1. Costello Formation macrobarite Δ¹⁷O values are closely clustered, with all analyses falling between −0.78‰ and −0.55‰, well below that of modern marine SO₄ [−0.09‰ (31)] and modern tropospheric O₂ (−0.516‰; ref. 30 and Fig. 2). δ³⁴S values in these samples cover a range that includes modern marine SO₄ (33) and overlap with previously published δ³⁴S CAS values from the Kasegalik Formation (Fig. 3A and ref. 42). Unlike δ³⁴S values, Δ³³S data significantly deviate from values for modern marine SO₄ [Δ³³S = +0.05‰ (33)] with predominantly negative values between −0.11‰ and −0.06‰ (Fig. 3A). Finally, Costello Formation macrobarite δ^138/134Ba data tightly cluster between +0.08‰ and +0.12‰, similar to values for modern and post-Marinoan marine barites (ref. 38 and Fig. 3B), but contrast with modern terrestrial, cold seep, or hydrothermal values (−0.49‰ to +0.52‰, −0.61‰ to +0.36‰, and −0.08‰ to −0.04‰, respectively; ref. 38).

Fig. 2.

Geochemical and GPP model results. (A) Triple-oxygen isotope results from the Costello Formation macrobarites (red circles) together with results from previously reported values (gray circles) that span the GOE and mid-Proterozoic (ref. 14 and references therein and ref. 46). (B) Box and whisker plots for low and high pO₂ scenarios during the GOE and low pO₂ post-GOE, showing the decline in GPP across the end-GOE. GPP fell dramatically across the end-GOE transition, even if pO₂ levels remained constant, and remained at very low levels through the mid-Proterozoic. Bottoms and tops of boxes correspond to 25th and 75th percentiles, respectively. Lower and upper whiskers correspond to 2.5th and 97.5th percentiles, respectively. The solid horizontal lines indicate medians, and dashed horizontal lines indicate means. Dotted vertical lines represent potential SO₄ ages based on different genetic models of barite formation.

Fig. 3.

Sulfur and barium isotope chemistry of Costello Formation macrobarites. (A) A δ³⁴S-Δ³³S cross-plot shows that Costello Formation macrobarites are consistently depleted in Δ³³S relative to modern seawater, while they overlap with modern values in δ³⁴S. δ³⁴S measurements also overlap with CAS measurements of the Kasegalik Formation (gray field; ref. 42). A comparison of syn- to post-GOE sulfate minerals shows that post-GOE samples are comparatively depleted in Δ³³S, possibly due to a change in sulfur inputs rather than microbial sulfur recycling. (B) δ^138/134Ba measurements of Costello Formation macrobarites compared with potential barite sources and post-Marinoan barites (38). The blue bar represents the typical analytical uncertainty on individual measurements. The range of values and small variance suggest a marine source of Ba for the Costello Formation macrobarites.

Discussion

Genesis of Costello Formation Barites.

The Costello Formation is the dominant host of barites (and Ba and SO₄) in the Belcher Group and we explore hypotheses for their genetic origin here. There is no stratigraphic, sedimentologic, mineralogic, or petrographic evidence for subaerial exposure in the Costello Formation (20–22), and thus it is unlikely that the macrobarites formed via direct precipitation of Ba- and SO₄-bearing salts in a strongly evaporitic environment. Rather, their large size, well-developed crystal habit, and lack of preferred orientation with respect to bedding strongly suggest that these barites formed postdepositionally rather than in the water column or at the sediment–water interface with coeval carbonates (Fig. 1). Since barite is an insoluble mineral (K_sp ∼ 10), significant diagenetic accumulations typically require the mixing of 2 separate fluids, one rich in Ba and the other in SO₄ (38, 40). The Ba- and S-isotope composition of barites thus constrains their respective source fluids and enables deduction of the significance of the Δ¹⁷O variations preserved therein.

We consider 5 possible sources of SO₄ to Costello Formation macrobarites. In stratigraphic order these are 1) remobilized SO₄ from the evaporitic and SO₄ pseudomorph-bearing horizons of the Kasegalik Formation, 2) remobilized SO₄ from the microbarite-bearing horizons of the Kasegalik Formation, 3) remobilized SO₄ from the formerly gypsum-bearing horizons of the McLeary Formation, 4) the ambient seawater SO₄ reservoir during the time of Costello Formation deposition, and 5) the downward movement of SO₄ from sediments overlying the Costello Formation. Two of these options (scenarios 2 and 3) can be ruled out on mass-balance grounds. The small stratigraphic expression of the microbarites in the Kasegalik Formation and their low abundance (∼1 ppm) make them a poor candidate source of SO₄ to the Costello Formation. Similarly, the McLeary Formation (ca. 1,945 Ma) was observed to have only a very small amount of chert-replaced gypsum. Scenario 5 is also highly unlikely because there are no known occurrences of SO₄-bearing minerals or their pseudomorphs in the stratigraphy overlying the Costello Formation.

The remaining 2 candidate sources of SO₄ set stratigraphically opposed temporal constraints on ∆¹⁷O signatures and, by consequence, Ba. A seawater source of SO₄ (scenario 4) that was penecontemporaneous with deposition of the Costello Formation would imply a more recent origin of the ∆¹⁷O signature, likely imparted ca. 1,930 Ma (assuming a constant sedimentation rate between age constraints), with a minimum age of 1,870 Ma ± 3 Ma from cross-cutting dykes and sills (Fig. 1 and refs. 20 and 43). A diffusive supply of seawater SO₄ to the barite-forming environment would also suggest that the global seawater SO₄ reservoir was characterized by large negative ∆¹⁷O values observed within the macrobarites, given that the Costello Formation bears no evidence for having formed in a strongly hydraulically restricted water mass. While this scenario is difficult to disprove sedimentologically, it would require abnormally low rates of microbial sulfur cycling to prevent the removal of ∆¹⁷O-anomalous SO₄ and abnormally high rates of continental SO₄ inputs to maintain a continuous supply of SO₄ that carried negative ∆¹⁷O anomalies; such a state has been inferred only for the exceptional oceanographic case of a meltwater lens following deglaciation of the Marinoan Snowball Earth (44). The alternative “remobilization” scenario 1, in which the SO₄ in Costello Formation macrobarites was derived from Kasegalik Formation SO₄-bearing minerals, places age constraints between 2,018.5 ± 1.0 Ma and 2,015.4 ± 1.8 Ma on ∆¹⁷O signatures. The Kasegalik Formation contains abundant chert-replaced gypsum pseudomorphs interspersed through some 25% of the stratigraphy (Fig. 1) and therefore, on the grounds of its sheer mass of SO₄-bearing minerals as well as its evaporitic setting, seems a likely source of SO₄ to the overlying Costello Formation. The similarity of δ³⁴S values of CAS measurements from the Kasegalik Formation (Fig. 3A and ref. 42) and those of the macrobarites of the Costello Formation cannot distinguish between these 2 possibilities, and additional evidence is required to discriminate between these sources of SO₄ for the Costello Formation macrobarites.

As noted previously (e.g., ref. 38), barites precipitated from large Ba reservoirs, such as seawater, generally exhibit a narrow Ba-isotope range close to the crustal average, whereas barites formed in more restricted closed-system settings show wide compositional ranges due to variable influences of inputs, diffusive transport, and precipitation–dissolution events (38). The narrow variance that was measured for δ^138/134Ba values for Costello Formation macrobarites (+0.10‰ ± 0.02‰; ±SD, n = 7) is indicative of a large Ba reservoir (e.g., ref. 39). We contend that this reservoir was contemporaneous seawater, since the mean δ^138/134Ba value of Costello Formation macrobarites closely matches those of ancient and modern marine precipitates (41). Due to the insoluble nature of barite, a contemporaneous seawater source of Ba implies a nonseawater source of SO₄ to the Costello Formation macrobarites. In turn, this constraint indicates that Costello Formation macrobarites record ∆¹⁷O anomalies inherited from SO₄ that were tens of millions of years older than the barites themselves (10), which is possible if SO₄-bearing fluids were stored in basinal or crustal brines (40, 45). That is, the remobilized SO₄ originated from the Kasegalik Formation (ca. 2,018.5 ± 0.5 Ma to 2,015.4 ± 1.6 Ma), whereas the Ba source—and thus barite precipitation—possesses a younger age of ca. 1,930 Ma (>1,870 Ma, based on cross-cutting dykes and sills).

A Productivity Crash to End the GOE?

∆¹⁷O values in the Costello Formation macrobarites (with sulfate that formed ca. 2,018.5 to 2,015.4 Ma) are similar to those deposited in the ca. 1,700-Ma Myrtle Shale Formation (14) and ca. 1,400-Ma Sibley Group (19), suggesting that the transition out of the GOE marks the onset of conditions that may have persisted across the >600-My interval separating these deposits (Fig. 2 and refs. 4, 14, 19, and 20). We set temporal constraints on this isotopic transition by comparing Costello Formation macrobarite results to the suggested youngest syn-GOE ∆¹⁷O values from the Tulomozero Formation in Russian Karelia (14, 46). The observation of highly positive δ¹³C values (up to +13.9‰) indicates that Tulomozero Formation sulfate was deposited during the LJE, placing an age of at least 2,108 to 2,057 Ma on these sediments and suggesting the shift in ∆¹⁷O values across the end-GOE occurred over <39 to 90 My. This illustrates the relative rapidity of the step change seen in the ∆¹⁷O record across the end-GOE.

The differences in ∆¹⁷O values between syn- and post-GOE samples could indicate a step change in the fraction of atmospheric O₂ preserved within SO₄ deposits (f_O2), a step change in the pO₂-pCO₂-GPP conditions that produced the atmospheric ∆¹⁷O anomaly, or some combination of these 2 factors. If the change were only due to f_O2, the shift in ∆¹⁷O values across the end-GOE transition would require a >4-fold increase in f_O2 after the GOE (from >0.02 to <0.15; SI Appendix, Fig. S3). However, if pO₂ was lower after the GOE interval (7), one would not expect more O₂ incorporation into SO₄ accompanying sulfide oxidation, thus placing the full burden on enhanced S cycling during the GOE to explain the ∆¹⁷O shift. Both syn- and post-GOE samples exhibit a similar range of δ³⁴S values (∼25‰; Fig. 3A) despite clear interformation variability. This is consistent with a similar degree of S cycling during both intervals and, by inference, no diminution of the syn-GOE ∆¹⁷O signal by enhanced O exchange with water (14, 46). Although the distinctly more positive ∆³³S values in syn-GOE samples could be interpreted to reflect greater S reoxidation at this time (35, 37), another explanation is that the lack of any covariation with δ³⁴S (cf. refs. 13 and 47) suggests that such signatures might instead reflect an isotopic shift associated with the oxidative weathering of “old” crustal sulfides (48).

In light of these considerations, we explore the ∆¹⁷O transition out of the GOE with respect to changing pO₂-pCO₂-GPP conditions. The spatial and isotopic consistency of the ∆¹⁷O results for the GOE (8 formations; 81 measurements; mean ∆¹⁷O = −0.18‰ ± 0.14‰, 2σ; refs. 14 and 46), as well as for the mid-Proterozoic (3 formations; 86 measurements; mean ∆¹⁷O = −0.68‰ ± 0.25‰, 2σ; refs. 14 and 19), is indicative of a step change in a global process as driving the transition. As a result, we use the Monte Carlo approach of ref. 19, along with independent estimates of pO₂, pCO₂, f_O2 (24), and other atmospheric parameters (ref. 19 and SI Appendix, Table S2), to interpret the ∆¹⁷O record primarily as a monitor of a state shift in GPP. We note that the interpretations presented here are conservative, as S recycling and reoxidation––if they were greater than assumed here––would drive ∆¹⁷O values toward 0 and thus GPP estimates toward higher values.

Across a range of reported pO₂ and pCO₂ conditions for the syn- and post-GOE intervals (SI Appendix, Table S3), our calculations show that a dramatic decrease in GPP characterized the end of the GOE. If high pO₂ levels were reached during the GOE (10% to 100% of modern pO₂; ref. 9), which then fell to 0.1% to 1% of modern in the post-GOE interval (7, 49), this transition could reflect a drop in GPP of nearly 200-fold, from a median value of over 1,100% of modern to a median value of ∼6% of modern (Fig. 2). This shift reflects the need to dilute a greater standing stock of tropospheric O₂ during the GOE. However, it is unclear whether syn-GOE biogeochemical cycles could support an oxygenic photoautotrophic biosphere that would achieve such high rates of carbon fixation and O₂ production (50, 51). The ∆¹⁷O record may not require such extreme pO₂ levels if their isotopic impact was tempered by some degree of enhanced S recycling and reoxidation during the GOE. A more likely scenario is a moderate transition, from initial pO₂ levels of 1% to 10% of modern to 0.1% to 1% of modern, which would have instead been accompanied by a reduction in GPP of ∼10-fold from a median value of ∼60% modern to 6% modern (SI Appendix, Table S1). Even in the unlikely case that pO₂ levels did not change across the end-GOE, the Δ¹⁷O values reported here still require over a 5-fold decrease in GPP across this transition (Fig. 2). Although our results indicate substantial drops in GPP are necessary to explain the Δ¹⁷O results, there is still a considerable range of possible Earth system states, underscoring the need for additional investigations of how biogeochemical cycles reorganized across the end-GOE transition.

Multiple studies have suggested a larger ocean–atmosphere oxidant inventory over the GOE interval compared with the following 1 billion y of Earth’s history (8, 12–18, 52). Interpretations of this decline in oxidant inventory are typified by falling marine SO₄ levels (12–14), as well as a potential decrease in organic carbon burial (50). The ∆¹⁷O evidence presented here suggests that a reduction in GPP may be a common underlying cause of the biogeochemical changes across this interval. This shift in GPP could reflect a change in the availability of critical nutrients, most likely phosphorus (50, 51, 53, 54), to sustain photoautotrophic growth. The switch from a nutrient feast to famine may have resulted from different Fe-P dynamics across this interval (51, 53), perhaps due to the exhaustion of weatherable apatite- and pyrite-rich sediments (8) as well as less P regeneration from biomass that had built up large nutrient reservoirs before the GOE (55). While the most conservative geological constraints limit the duration of this Earth system “tipping point” to <200 My, the geochemical, geochronological, and sedimentological observations reported here point toward a much quicker timescale (<39 to 90 My; Fig. 2) for the end-GOE transition. Cumulatively these findings suggest the end-GOE transition was potentially one of the largest sustained shifts in the productivity of the biosphere, rivaling the colonization of the terrestrial realm by land plants [>2-fold (56)] and the Permo-Triassic mass extinction [>2-fold (32)] and perhaps even approaching the advent of oxygenic photosynthesis [∼1,000-fold (57)] in magnitude.

Conclusion.

We present data suggesting that the end-GOE transition marks one of the most pronounced sustained changes in the productivity of the biosphere across all of Earth’s history. Moreover, our results strengthen the inextricable link between the ultimate source of oxygen production (the marine biosphere) and the oxidation of the Earth’s surface environment. We find a drop in GPP >5-fold, but possibly as much as 2 orders of magnitude across the end-GOE transition using a Monte Carlo approach paired with published estimates of pO₂ and pCO₂. This drop was likely brought about by a large decrease in nutrients supplied to the biosphere that, in turn, marked the conclusion of the GOE and ushered in the subsequent 1-billion-y interval characterized by markedly low and stable GPP compared with the modern Earth. Although the end-GOE is not considered a major biotic event, our results show that the decrease in gross primary productivity across this transition eclipses even the largest extinction events in all of Earth’s history.