Introduction

The central targets for reconstructing Earth’s ancient atmosphere and climate are pCO2 and pO2 as they intimately relate to Earth’s habitability and global metabolism1. Traditionally, paleo-atmospheric composition is indirectly inferred via stable isotope reconstructions2,3,4 or through the directionality of change in redox sensitive metals and their isotopes in marine sediments5,6,7. As the diversity of approaches to constrain atmospheric chemistry has increased, so too has our uncertainty of the timing of Earth’s rise to modern-like pO2 levels. These estimates span over 500 million years, a wider range than constraints on the original Great Oxidation Event ca. 2500–2200 Ma8,9,10,11,12,13. This uncertainty in timing has prevented the isolation of the mechanism(s) that drove oxygenation (e.g., tectonics, glaciation, or marine/terrestrial biology) and its ensuing redox implications, both of which are central to our understanding of the Earth system.

A more direct and emerging approach employing the triple oxygen isotope composition (Δ’17O) of sulphate minerals was recently proposed as a sharper quantitative tool for ancient atmospheric reconstructions14. It is known from observations in the modern atmosphere that tropospheric O2 carries a large, anomalous, and negative Δ’17O signature15. The magnitude of the isotope signature in atmospheric O2 scales with pO2, pCO2, and gross primary production (GPP) as a result of atmospheric chemistry and overall isotope mass-balance15. It is well-documented in the Proterozoic (2500–539 Ma14,16,17) that sulphate minerals carry a large, anomalously negative Δ’17O composition that is most likely derived from contemporaneous tropospheric O2. Existing Ediacaran measurements are difficult to assign to a marine evaporite basin since samples are from salt-plugs and nodular sulphates from Russia, which may have complex (non-marine) histories. The closest, unequivocally marine evaporites are found in Northwestern Canada and Victoria Island 820–750 Ma (18 and refs. therein). Unlike Proterozoic sulphate records, the Δ’17O of Cretaceous–Cenozoic marine barite (BaSO4)19, as well as studies of modern seawater20 and riverine21,22,23 sulphate, show no resolvable atmospheric O2 fingerprint. These geological sulphate catalogues instead record biogeochemistry — specifically, microbial sulphate reduction (MSR)19,20,22,24. Today, it is thought that MSR partially resets the oxygen isotope composition of marine sulphate to an isotopic equilibrium with environmental H2O19,20,23,25, and further that terrestrial pyrite oxidation (and sulphate generation) does not adopt a resolvable atmospheric signal22. At face value then, Earth’s sulphur and oxygen cycles must have changed significantly sometime between ~800–130 Ma to preclude the incorporation of atmospheric O2 signatures into marine sulphate.

Generally, during pyrite oxidation, Fe3+ (and to a lesser extent, O2) steals electrons from pyrite to generate aqueous sulphate and Fe2+26,27. The rate of this reaction is largely a function of oxidant availability, which in turn is related to diffusion and transport of electron acceptors to the pyrite surface28. Numerous (labeled and unlabeled) isotope studies have tried to constrain the sources of oxygen atoms in the sulphate produced from pyrite oxidation over a range of temperature, pressure, pH and (a)biological conditions (see SI section 5). With heterogeneous results, these experiments commonly observe that sulphate obtains 0–30% of its O atoms from O2, with the remaining O atoms donated from H2O29,30,31,32,33,34,35,36. At a more granular level, the capacity to directly incorporate an O atom from O2 into sulphate is a function of the local availability of Fe3+30, whereby elevated iron contents and vigorous iron redox recycling limits the direct incorporation of O2 into sulphate. This leaves an obvious role for iron cycling in affecting the isotopic composition of sulphate across Earth history.

Results and Discussion

Here we present 189 new high-precision37 triple oxygen isotope (δ18O, Δ’17O) measurements from twelve heavily sampled, globally distributed and volumetrically significant marine sulphate evaporite basins spanning the last 542 million years of Earth history (Fig. 1, Supplementary Data 1, and Tables S2 & S3 for more information). We treat each evaporite basin, sampled with between 5 and 49 measurements, as representative of marine sulphate at the time of deposition. This is supported by the fact that the contemporaneous basins yield statistically indistinguishable oxygen isotope compositions (e.g., Canadian Prairie Formation and the Iowa Pinicon Ridge Formation, see Table S3). Each basin carries some internal variability in both δ18O and Δ’17O associated with basin restriction, depositional timescales, and secondary processes (see24 and SI for calibration, correction factors, and extended statistical analyses). However, the most significant feature of the data set is a step change in Δ’17O in the mid-Paleozoic, supported by a Wilcoxon rank sum test (p < 0.05). The basins ≤387.7 Ma (n = 9 with this study, n = 3 from scaled published data16,24, see SI) are overlapping in both Δ’17O and δ18O (with interquartile ranges of 0.2 to 3.3‰ for δ18O and 0.01 to 0.06‰ for Δ’17O), carry near-zero Δ’17O compositions and overall lower δ18O values (see Table S2, Fig. 2). Conversely, eight basins between 1050 and 420 Ma (n = 3 with this study, n = 5 from scaled published data18) have overall higher and more variable δ18O and significantly more negative Δ’17O compositions (see Table S2, Fig. 2). A two-sample t-test finds that the younger (≤387.7 Ma) and older (≥420 Ma) Δ’17O populations are distinct and statistically offset by between 0.08 and 0.10‰ (as defined by the 95% confidence interval). These two populations of data capture a state change in the composition of marine sulphate, which requires a transition in the interaction of the sulphur and oxygen cycles between 420 and 387.7 Ma.

Fig. 1: A billion-year record of the oxygen isotope composition of marine sulphate.
Fig. 1: A billion-year record of the oxygen isotope composition of marine sulphate.The alternative text for this image may have been generated using AI.
Full size image

The oxygen isotope composition (Δ’17O) of volumetrically significant marine sulphate evaporite basins over approximately the last 1050 million years. Filled (this study) and open circles (published data16,18,24 scaled, see SI) are represented as a function of samples per basin, centered on the median value, and with error bars that denote the 25th and 75th quartiles of each distribution. A smoothed fit, with a 95% confidence interval for the Cenozoic-Cretaceous sulphate record is noted in the filled blue-green region19. At the top is a timeline of geologic events that have been linked to a rise in atmospheric pO2. The two Neoproterozoic Snowball Earth events (Cryogenian Glaciations) are represented by vertical blue bars. Data from Marinoan barite crystal fans16 (i.e., not evaporite deposits) is excluded. The gray dashed bars represent the temporal constraint on pO2 rise based on the marine sulphate evaporite record.

Fig. 2: Triple oxygen isotope composition of marine sulphate minerals.
Fig. 2: Triple oxygen isotope composition of marine sulphate minerals.The alternative text for this image may have been generated using AI.
Full size image

The triple oxygen isotope composition (Δ’17O against δ18O) of marine sulphate evaporite deposits (filled circles=this work; open circles=literature16,18,24; marker size scales with sample set size). Each circle reflects the median basinal value, with error bars reflecting the 25th and 75th percentiles. As in Fig. 1, blue-green symbols reflect basins ≤387.7 Ma, whereas brown symbols are ≥420 Ma. The light blue-green regression line reflects the Cenozoic-Cretaceous marine barite record19. The two triangles in the lower right are sulphate isomer equilibrium with seawater at 15 °C23. The long-dashed line — approximated as linear — represents sulphate that derives between 0 to 25% of its O atoms from (modern) atmospheric O2 and the remainder from seawater H2O. Finally, a suite of vectors below the key indicates the directionality of change if (i) meteoric waters are involved, (ii) if paleo-O2 had a more depleted Δ’17O composition, or (iii) if sulphate is overprinted via thermodynamic equilibria with seawater.

Possible Drivers of Sulphate Oxygen Isotope Change

The timing of the transition in the Δ’17O composition of marine sulphate is later than most events thought to drive significant changes in atmospheric pO2. These events include the Neoproterozoic Snowball Earth glaciations16,38, the origin of animals39, the onset of vigorous marine bioturbation40,41, inferred changes in Ediacaran ocean redox conditions39, and the initial deposition of large, bedded evaporite deposits42. All of these events, separated by >300 million years, have been argued to be drivers or products of changes to the inventory of atmospheric O2. However, given that marine evaporites formed between 420 Ma and 1050 Ma are isotopically consistent and distinct from younger units, it is now apparent that none of these triggers or events ushered in a change in O2 significant enough to be resolved in the Δ’17O sulphate record. Instead, the loss of negative, anomalous Δ’17O signatures coincides with the proliferation of vascular land plants and coincident change in terrestrial soils43,44. The question then becomes ─ how did Earth’s major geochemical cycles (notably O and S) respond to this paleobiological innovation?

Since the Great Oxidation Event45,46, tropospheric O2 would have been the most significant carrier of a negative Δ’17O anomaly15,47. As such, the negative Δ’17O preserved in sulphate deposits ≥420 Ma almost certainly preserves some fraction of contemporaneous atmospheric O2 (Figure S7). The loss of the Δ’17O anomaly between 420 and 387.7 Ma requires either a change in sulphate generation pathways to preclude the entrainment of oxygen from O2, or a more vigorous S cycle capable of erasing a terrestrial O isotope anomaly via enhanced riverine or marine microbial sulphate recycling20. Fortunately, the rich sedimentary record of δ34S of pyrite helps inform this question. A statistical analysis of compiled δ34S data illustrate that the mean offset between sulphate and sulphide is statistically identical from 500–300 Ma40,48. Any significant change in MSR activity and a change in the relative importance of pyrite burial would manifest in the δ34S record, suggesting that the more likely trigger in the Δ’17O signal is from terrestrial environments and pyrite oxidation itself.

A Transition in Pyrite Weathering

In the modern, rooted vascular plants deepen the locus of terrestrial pyrite weathering, physically and chemically breaking down substrates, and stabilizing a meter(s)-thick soil mantle49. Beneath the soil layer, pyrite weathering reactions proceed within bedrock fractures28. This environment is, in most cases, physically dislocated from contact with atmospheric O2, leaving diffusion responsible for the addition and subtraction of reactants and products at the literal site of pyrite oxidation28. Diffused O2 directly or indirectly catalyzes complex Fe–S–O redox cycling, with Fe3+, S intermediates, and oxygen species all serving as potential oxidants for pyrite. The consequence of the suite of oxidants in these environments is no resolvable isotopic relationship to modern atmospheric O222,23. This is consistent with modern river systems, which are dominated by sulphate oxygen isotope compositions most easily explained via isotopic equilibrium with local waters22.

Similar to modern rivers and Cenozoic–Cretaceous barite19,22, the ≤387.7 Ma marine evaporite record preserves a signal more related to equilibrium with water and MSR (see annotations on Fig. 2) than the addition of tropospheric O2. This suggests that a modern-like weathering regime was active in the early Devonian, making modern river systems, for example, a suitable proxy for the last 387.7 million years; the lack of mass-independent Δ’17O in sulphate is thus the expectation. In the early Paleozoic and throughout the Proterozoic, however, the isotope compositions of sulphate evaporites fall more toward vectors describing atmospheric O2 incorporation (Fig. 2, dashed line; Figure S7). Excluded from the present discussion are the post-Marinoan glaciation barite crystal fans that record large, anomalous oxygen isotope depletions16. Although sulphate minerals, the barite crystal fans are not fit for direct comparison to evaporite records, being sedimentologically distinct, yet their interpretation may benefit from many of the arguments herein. In a world before soils, the locus of terrestrial pyrite oxidation would be shallower in the regolith43, if not also drier4. These features would limit the capacity of Fe recycling to physically keep O2 at a distance from pyrite and, as such, allow the incorporation of atmospheric O2. This is all in keeping with increasing water-rock interactions50 and the generation of thick soils with secondary mineral phases in plant-influenced environments51. It also aligns with the experimental data that show, under some specific conditions, pyrite oxidation can indeed incorporate oxygen atoms from O230,32,52.

Rise in Atmospheric Oxygen Fueled by Land Plants

The entire infrastructure on which the Δ’17O proxy is predicated ties the magnitude of the Δ’17O anomaly in sulphate to pCO2, pO2, and GPP. The fraction of sulphate oxygen derived from atmospheric O2 is largely assumed to be invariant and often fixed at a range of 8–15%. Previous forward53 and inverse3,54 models of Paleozoic atmospheric change vary widely, but most call on a dramatic mid-Paleozoic increase in pO2/pCO2 even before considering changes in GPP related to land plants (see SI for sensitivity tests, model outputs/predictions and expanded discussion). This modeled increase in pO2/pCO2 drives a predicted increase in the sulphate Δ’17O, similar in direction to the observed evaporite record (Fig. 1). That said, if changes in pO2 contributed to increased Δ’17O, that oxygenation event would have to occur in the early Devonian. Instead, we propose that slowing or even stopping the capacity to incorporate O2 into sulphate—making the fraction of O from O2 negligible—is a simpler and better fit to the data. If correct, the early Devonian may mark the end of the capacity of marine evaporites to record paleo-atmospheric compositions.

The energy required for oxidative weathering is ultimately linked to the atmospheric overburden of pO2. This requires enough molecular oxygen present by 420–387.7 Ma to power modern-like pyrite oxidation. This interval, the latest Silurian to Early Devonian, also captures the development of “true roots” in vascular land plants (e.g., ~407 Ma Rhynie Chert Lagerstätten) and generally deeper rooting systems—up to 1 m in depth44,55. We hypothesize that the coincident shift in the Δ’17O composition of evaporites reflects a deepening of the global terrestrial weathering environment associated with these early rooting systems by at least 387.7 Ma. This argument supports the hypothesis that the proliferation of land plants took Earth to a modern-like S cycle with putatively modern-like levels of pO23,56, perhaps through the increase in the C:P ratio of terrestrial biomass (e.g.57,) and enhanced terrestrial organic carbon burial associated with the Devonian proliferation of vascular plants. It also follows that the pO2 levels necessary to trigger the change in the Δ’17O signal were not crossed in the Neoproterozoic or early Paleozoic, as those evaporite Δ’17O signatures look more like the Proterozoic than Mesozoic–Cenozoic (see Figure S7). These findings allow for the refinement of models targeting the relationship between land plant evolution, GPP and atmospheric pO2 (3, see SI for model simulations).

As marine sulphate stands as the largest oxidant pool in the modern ocean, its associated evaporite record provides a critical window into Earth’s evolving redox budget. In reconstructing the last ~1000 million years of seawater sulphate, the mid-Paleozoic holds a key transition—one from a world where atmospheric oxygen was commonly found in marine sulphate to an environment recording only mass-dependent biogeochemical cycling. This transition is coincident with the development of thicker terrestrial soils and the advent of rooted vascular plants. These changes would have driven pyrite oxidation deep into waterlogged environments, where a complex oxygen-driven Fe–S redox cycle serves to generate sulphate without O2, just like the modern. We argue that a contemporaneous rise of pO2, presumably associated with an increase in terrestrial organic carbon burial and consistent with the data presented herein (see Figure S8), drove the changes in terrestrial Fe–S cycling. Notably, this evidence for oxygenation between 420 and 387.7 Ma significantly postdates calls for a Neoproterozoic8 and even a putative end-Ordovician58 increase in pO2 and broadly agrees with recent evidence for a sustained increase in pO2 across the Paleozoic59. Further, the persistent preservation of anomalous sulphate throughout the Proterozoic and early Paleozoic implies that even if present, earlier oxygenation events never reached mid-Paleozoic levels of pO2. We find that the proliferation of vascular plants and arboresence may be the root cause of Earth’s ultimate rise in pO2, linking evolutionary advancements preserved in fossil records to a critical transition in Earth’s surface redox state. As the loss of anomalous sulphur isotope signatures has come to be the hallmark of Earth’s first major oxidation45, the loss of anomalous oxygen isotope signatures may mark the second and final rise of pO2.

Methods

Isotope Nomenclature

Standard definitions for δ18O and Δ’17O are as follows:

$${{{\rm{\delta }}}}{}^{18}{{{\rm{O}}}}_{{{{\rm{sample}}}}}={10}^{3}{\cdot }\left(\frac{{}^{18}{{{\rm{R}}}}_{{{{\rm{sample}}}}}}{{}^{18}{{{\rm{R}}}}_{{{{\rm{standard}}}}}}-1\right)$$
(1)

and

$$\Delta {}^{{\prime} 17}{{{\rm{O}}}}_{{{{\rm{sample}}}}}={10}^{3}{\cdot }\left({{\mathrm{ln}}}\left(\frac{{{{\rm{\delta }}}}{}^{17}{{{\rm{O}}}}_{{{{\rm{sample}}}}}}{1000}+1\right)-{{{{\rm{\theta }}}}}_{{{{\rm{RF}}}}}{\cdot }{{\mathrm{ln}}}\left(\frac{{{{\rm{\delta }}}}{}^{18}{{{\rm{O}}}}_{{{{\rm{sample}}}}}}{1000}+1\right)\right)$$
(2)

Here, θRF is the slope of the mass dependent reference line, which we define as θRF = 0.5305. Using a set of internal standards, we report a precision on sulfate of ±0.02‰ (1σ) for Δ’17O. All δ18O values are reported with respect to VSMOW using a scale and drift correction using IAEA-SO5 and IAEA-SO6. The Δ’17O data is presented on the silicate-derived VSMOW/SLAP scale described below.

Sample Treatment

There are four main stages in the preparation and analysis of the Δ’17O of natural sulphate samples: (1) chemical purification, (2) conversion to analyte gas using a fluorinating agent, (3) analyte gas purification protocols, and (4) data treatment and standardization. (1) Two methods were used to purify sulphate in this study: a) ion exchange resin chromatography60 and b) dissolution in a chelating solution of diethylenetriamine pentaacetate (DTPA)61. Both methods precipitate BaSO4 from purified SO42- solutions and yield isotopically indistinguishable results. (2) BaSO4 conversion to analyte O2 was performed by laser heating under an F2 atmosphere37,62. This method results in non-quantitative O2 yields. Yields at Harvard University are on average 33 ± 8% (full range from 11–58%) for F2 fluorination and sample gas cleaning and are reported with all published data such that data can be corrected. (3) Analyte O2 gas was purified via a series of cryogenic and chromatographic steps prior to isotopic analysis37. We ensured that nitrogen-containing compounds were removed prior to analysis, to prevent isobaric interferences with the 17O-containing m/z 33 beam63. (4) All data here (standards and unknowns) are reported on a VSMOW/SLAP scale using a three-point silicate-air reference frame (UWG-2: δ18O = 5.70‰, Δ’17O = −0.085‰; NBS 28: δ18O = 9.58‰, Δ’17O = -0.083‰; air O2: δ18O = 24.05‰, Δ’17O = −0.501‰) with accepted values64.

Inter-Lab and Inter-Method Data Comparison

Data in this study and from previous work16,18,19,24 has been scaled for direct comparison. Due to isobaric interferences, the measurement of a triple-oxygen isotope composition in sulphate is performed by converting SO42− to O2, the measured analyte gas. Several methods exist to achieve this conversion (e.g.37,65,66,), and these are not uniform between labs and datasets, nor are published standardizations. Methods such as laser fluorination (used in this study) that do not generate 100% yield of SO42−–O inherently introduce greater uncertainty. It is important to note that as new labs and new methods65,66 become available, rigorous inter-lab and inter-method comparisons are being undertaken66.