Skip to main content
NCBI home page
Nature Communications logoLink to Nature Communications
. 2026 Aug 27;17:10294. doi: 10.1038/s41467-026-76597-y

A nutrient control on oxygenation dynamics during Earth’s Great Oxidation Episode

Lewis J Alcott 1,2,✉, Benjamin J W Mills 2, Andrey Bekker 3,4, Zidong Peng 2,5, Simon W Poulton 2
PMCID: PMC13620156  PMID: 42805977

Abstract

Earth’s first rise in atmospheric oxygen occurred over a protracted period from ca. 2.43 to 2.06 billion years ago during the Great Oxidation Episode (GOE). The GOE was marked by extreme climatic instability, which likely impacted the bioavailability of the principal limiting nutrient, phosphorus. Here, we investigate phosphorus cycling across the final two glaciations of the GOE, as documented by the Transvaal Supergroup, South Africa. Utilising a phosphorus phase partitioning approach, we show that following an enhanced weathering-induced phosphorus influx that stimulated oxygenation in the aftermath of glaciation, sedimentary phosphorus recycling was controlled by feedback-driven fluctuations in sulphate availability and oceanic redox state. Biogeochemical modelling suggests that this dynamic interplay drove major instability in atmospheric oxygen levels during Earth’s protracted transition to a persistently oxygenated atmosphere.

Subject terms: Element cycles, Geochemistry

Earth’s first rise in atmospheric oxygen was marked by extreme climatic instability. Phosphorus phase partitioning and biogeochemical modelling show that feedbacks between nutrient availability, redox and deglaciation controlled oxygenation.

Introduction

The Great Oxidation Episode (GOE) documents the transition to a permanently oxygenated atmosphere1, and represents a defining interval in Earth history with respect to the subsequent evolution of seawater redox chemistry and the role of marine oxygenation in the course of biological evolution2–4. While multiple independent lines of evidence support an overall rise in atmospheric oxygen across the GOE4, the loss of mass-independent fractionation of sulphur isotopes (MIF-S) provides the most precise tool for assessing the timing and dynamics of atmospheric oxygenation1,5–12. Indeed, based on improved, geochronologically calibrated multiple sulphur isotope records, the onset of the GOE has been constrained to ca. 2.43 billion years ago (Ga)6,7, which represents the initial rise of atmospheric oxygen to above 10−6 PAL (Present Atmospheric Level)13,14.

The overall transition to higher atmospheric oxygen levels is linked to extreme perturbations in global climate1,6,15. Global lithostratigraphic correlations infer multiple glacial intervals during the GOE, with the widespread, low-latitude extent of earlier glaciations implying possible ‘Snowball Earth’ conditions, prior to the final, more regional glaciation at ca. 2.25 Ga6,16–18. A strong interplay between these glacial events and fluctuating oxygen levels is apparent, with oxygenation peaking during non-glacial intervals1,6,9,19,20. While this oscillatory behaviour likely occurred on a global scale, possible spatial heterogeneity in the MIF-S signal highlights that its manifestation in the marine geological record is complex11,12. Going further, the recognition of oscillatory oxygen levels from MIF-S signals is further complicated by potential detrital contributions and post-depositional overprints21,22. Nevertheless, recent work has demonstrated that the aftermath of the final widespread glaciation marks the onset of coupled atmosphere-ocean oxygenation23, as indicated by a clear coupling between lower thallium isotope ratios (documenting ocean oxygenation) and the disappearance of MIF-S (documenting atmospheric oxygenation)1.

While the precise nature of oceanic redox conditions and their dynamics across the GOE remains poorly resolved, it is clear that deeper waters remained predominantly ferruginous (anoxic and Fe2+-bearing)2, with spatially and temporally restricted euxinia (anoxic and sulphidic water-column conditions) occurring in shallow waters, at least in the aftermath of the final widespread glaciation1. These redox dynamics would have greatly impacted the bioavailability of phosphorus (P), the ultimate limiting nutrient on geological timescales24, which is thought to have been fundamental in priming and driving the GOE25,26.

Under ferruginous conditions, there is potential for extensive drawdown and fixation of P through uptake in association with iron minerals27,28. However, sulphate and organic carbon availability may diminish this iron trap through the production of H2S during early diagenesis, leading to the reductive dissolution of iron (oxyhydr)oxide minerals and release of associated P25,29,30. This release of P to pore-waters and seawater is commonly enhanced by the preferential release of organic-bound P (Porg) relative to organic carbon (Corg) during anaerobic remineralization, leading to elevated molar Corg/Porg ratios relative to the Redfield ratio of 106:131,32. After release into sediment pore-waters, P may be fixed via ‘sink-switching’ to authigenic phases, such as carbonate fluorapatite (CFA)33,34 or vivianite34–36. However, under anoxic water-column conditions, particularly when the water column is euxinic, or when pore-waters are sulphidic close to the sediment-water interface (which removes the Fe-(oxyhydr)oxide trap for P), P can be extensively recycled back to the water column, where it may stimulate further productivity, leading to a net increase in oxygen production25,30,37. However, these redox-promoted biogeochemical processes have not been investigated for any part of the GOE, and hence nutrient controls on dynamic atmospheric oxygen levels across the GOE remain entirely unconstrained.

To investigate the role of P cycling in driving oxygenation dynamics across the GOE, we focus on shale samples in well-preserved drill-core from the ca. 2.32-2.25 Ga Rooihoogte and Timeball Hill formations of the Transvaal Supergroup, South Africa. This succession documents clear atmosphere-ocean redox changes across the final widespread glaciation and subsequent regional glaciation of the GOE1,23,38, and we build upon existing redox data with new redox-sensitive trace element (RSTE) analyses, alongside a detailed study of P phase partitioning. In combination with biogeochemical modelling, this approach demonstrates how feedbacks between P cycling, oceanic redox conditions and climate drove major instability in oxygenation during Earth’s protracted transition to a persistently oxygenated atmosphere.

Results and discussion

Atmospheric and oceanic redox conditions

Our new RSTE (shown as Re/U, Re/Mo and Mo/U ratios39) and P phase partitioning data for drill-core EBA-2 (see Methods and Supplementary Information for a full description of the geological setting, analytical techniques and all data) are presented alongside existing multiple S isotope and Corg concentration data1 in Fig. 1. The Rooihoogte and Timeball Hill formations in drill-core EBA-2, which were deposited in a shallow-marine, deltaic setting with unrestricted connection to the ocean40,41, capture the nature of oxygenation in the aftermath of the final widespread glaciation, while the overlying Reitfontein Diamictite documents the subsequent regional-scale glaciation42 (Fig. 1).

Fig. 1. Geochemical and isotopic profiles through drill-core EBA-2.

Fig. 1

Open in a new tab

Trends in atmospheric oxygenation and ocean redox conditions are based on Fe-S1 and redox-sensitive trace elements (this study) systematics. Red lines on U/Al, Mo/Al and Re/Al plots represent average ratios for upper continental crust49. Multiple sulphur isotope data1 are included to emphasise the atmospheric oxygenation history, and the Δ33S record also includes data from an additional drill core (EBA-1). Red lines on Corg:Porg and Corg/Preac plots denote the Redfield ratio (106:1). Corg – organic carbon, Porg – organic bound phosphorus, Preac - reactive phosphorus. Intervals (A–E) correspond to broad atmosphere-ocean redox intervals1, but we note that even higher-frequency redox fluctuations may have occurred within these broad intervals12. The third glacial interval of the Great Oxidation Episode (GOE) is older than the Rooihoogte Formation. Roo.- Rooihoogte Formation, Riet. – Reitfontein Diamictite in the Timeball Hill Formation, Bosh. - Boshoek Formation.

Large MIF-S signatures indicative of a functionally anoxic atmosphere in the lower Rooihoogte Formation are initially interrupted by a zone of mass-dependent S isotope (MDF-S) values that have been suggested to reflect a transient interval of atmospheric oxygenation1,10, prior to a second apparent loss of the MIF-S signal across the Rooihoogte-Timeball Hill Formation boundary (Fig. 1). This second loss of the MIF-S signal has traditionally been considered to document the permanent rise of atmospheric oxygen8,10. However, the multiple S isotope record implies that fluctuations between a functionally anoxic and oxic atmosphere may have persisted through the overlying strata, with the permanent transition to an oxygenated atmosphere occurring in association with the Lomagundi C isotope excursion at ca. 2.22 Ga1. Nevertheless, while we consider the nature of P controls on atmospheric and ocean oxygenation throughout the section, our main focus is on the larger-scale ocean redox and nutrient interactions that occurred in the immediate aftermath of glacial melting, lower in the section.

Local, shallow-marine redox fluctuations through the section are closely linked to the dynamics of atmospheric oxygenation (Fig. 1; for simplicity, we divide the section into five intervals (A-E) based on the dominant atmospheric and oceanic redox state1). While our RSTE data are entirely consistent with redox reconstructions based on Fe-S systematics1 and lower-resolution RSTE (U and Mo) data23, our specific utilisation of RSTE ratios39 provides additional nuance due to the enhanced ability to distinguish dysoxic from oxic bottom-water conditions (see Supplementary Information Figs. S2–4). In brief, U sequestration commonly occurs under anoxic bottom-water conditions43, while Re is more sensitive to oxygen depletion and begins to be sequestered under dysoxic conditions44,45. By contrast, substantial drawdown of Mo either requires anoxia and the presence of dissolved sulphide, or a particulate shuttle mechanism whereby Mo (but not U or Re) is sequestered due to uptake by Fe minerals precipitating at the redoxcline above deeper ferruginous waters46–48. Therefore, high Re/U and Re/Mo values commonly indicate dysoxic conditions (further supported by high Re/Al and low highly reactive (FeHR) to total Fe (FeT) values), whereas low Re/Mo and high Re/U values commonly develop under anoxic, ferruginous conditions (supported by high Re/Al, U/Al and FeHR/FeT, and low pyrite (Fepy) to FeHR values)37. Euxinic conditions are commonly characterised by high Mo/U values, alongside high FeHR/FeT and Fepy/FeHR values (also supported by high Re/Al, U/Al and Mo/Al, and low Re/Mo and Re/U values)37.

Utilising this framework alongside Fe-S systematics1 (Figs. S2, 3) provides evidence for fluctuating redox conditions under an anoxic atmosphere in interval A (Fig. 1). Specifically, oxic depositional conditions are indicated by low Re/U and Re/Mo values (Fig. 1), alongside generally low FeHR*/FeT values (where FeHR* includes a correction for loss of FeHR to sheet-silicate minerals during diagenesis1,2; see Supplementary Information). Occasionally, elevated FeHR*/FeT and FeT/Al values do occur, however, with some samples having elevated Fepy/FeHR values (Figs. S2, 3). Nevertheless, Re/Mo and Re/U values remain low (Fig. 1), suggesting that Fe enrichments likely occurred due to periodic upwelling of ferruginous waters into oxic shallow-waters, with elevated Fepy/FeHR values likely reflecting diagenetic pyrite formation, which is consistent with relatively heavy δ34S pyrite values (Fig. 1). Thus, interval A appears to have experienced fluctuating seawater redox conditions linked to periodic upwelling.

In interval B, where multiple S-isotope systematics support atmospheric oxygenation1,10, the available ocean redox data are limited, but initially suggest oxic depositional conditions (Fig. S3; see also ref. 1 for supporting data from a second drill-core), followed by the development of euxinia towards the top of the interval (indicated by Fe-S systematics, elevated Mo/U ratios, and co-variation in Mo and U enrichment factors; Fig. 1 and Figs. S3, 4). Euxinia persisted as the atmosphere transitioned back to an anoxic state through interval C and into the basal part of the Timeball Hill Formation, before the development of ferruginous conditions (Fig. 1 and Fig. S3). This short-lived ferruginous interval brackets the initial stage of the second period of atmospheric oxygenation, before the atmosphere again transitioned to an anoxic state (Fig. 1).

Following the initial post-glacial redox dynamics, low FeHR*/FeT ratios (Fig. S2), combined with generally low Re/U and Re/Mo ratios (Fig. 1), indicate predominantly oxic shallow-marine conditions through interval D, although the occurrence of occasionally elevated Re/U ratios suggests the periodic development of dysoxic conditions in bottom-waters or at the sediment-water interface39. Through interval E in the upper Timeball Hill Formation, fluctuations in FeHR*/FeT, Re/U and Re/Mo ratios (Fig. 1 and Fig. S2) indicate variability in water-column redox conditions (from ferruginous to dysoxic/oxic), prior to the development of more persistent ferruginous conditions leading up to, and during, the final regional glaciation (Fig. 1 and Fig. S2).

Phosphorus dynamics

Phosphorus concentrations throughout much of the section (expressed as P/Al; Fig. 1) are commonly lower than upper continental crust values (0.009)49, with the exception of a notable zone of elevated values across the Rooihoogte-Timeball Hill formation boundary, and a few samples in the upper Timeball Hill Formation. These Rooihoogte-Timeball Hill formation boundary samples appear to document extensive P drawdown in association with organic matter (Fig. 1), highlighting a productivity control on P sequestration. However, the extent to which P is retained in the sediment during early diagenesis also affects the P loading of the sediments31,32,50. To assess this, we consider redox controls on Corg/Porg and Corg/Preac ratios, where Preac represents potentially bioavailable P (comprising organic, authigenic and Fe-bound P; see Supplementary Information).

We first note that it is not simply the dominant oceanic redox state that determines whether P is recycled back to the water column (indicated by Corg:Porg and Corg:Preac ratios above 106:1) or retained in the sediment. Rather, it is the generation of dissolved sulphide during very early diagenesis that exerts a major control, which would itself have been affected (at least in part) by low (and fluctuating) sulphate concentrations prior to persistent oxygenation of the atmosphere25. As a first-order assessment of relative sulphate availability, we consider pyrite S isotope (δ34S) values (Fig. 1).

Although there are a variety of potential controls on pyrite δ34S values in ancient sediments4,51,52, very low compositions (relative to the earlier Precambrian4) of about −35‰ across the Rooihoogte-Timeball Hill formation boundary (Fig. 1) are consistent with elevated sulphate concentrations (above 200 μM53). In considering alternative possibilities, we note that there is no sedimentological evidence for a change in sedimentation rate that might influence pyrite δ34S values54, and instead, the pyrite δ34S trends occur coincident with major changes in atmospheric oxygenation and ocean redox conditions (Fig. 1). This suggests an enhanced flux of sulphate to the oceans (which increases δ34S fractionation in pyrite51,55) driven by oxidative weathering of the continents linked to rising atmospheric oxygen8. We thus distinguish our P phase partitioning data based on the sulphur isotope composition, as a reflection of broad changes in seawater sulphate availability driven by differing intensity of oxidative continental weathering during fluctuating atmospheric oxygen levels and associated oceanic redox state (Fig. 2).

Fig. 2. Relationship between organic carbon and different phosphorus pools for samples from drill-core EBA-2.

Fig. 2

Open in a new tab

A Corg:Porg; B Corg:Preac. Black dashed lines represent the Redfield Ratio (106:1) and the blue dashed line represents an inferred Corg:Porg ratio of 300:1 for early Precambrian sediments deposited under highly oligotrophic, P-limited conditions27. Symbols represent defined redox conditions as described. Colour of symbols denotes δ34S values. Corg – organic carbon, Porg – organic bound phosphorus, Preac - reactive phosphorus.

This approach clearly highlights that samples deposited during time intervals of probably increased seawater sulphate concentrations (dominantly corresponding to those at the Rooihoogte-Timeball Hill Formation boundary; Fig. 1) generally have higher Corg/Porg and Corg/Preac ratios, regardless of the precise redox state of the water column (Fig. 2). The Corg/Porg and Corg/Preac ratios for these samples are considerably elevated above both the Redfield ratio and an assumed ratio of 300:127 for highly P-limited conditions in globally ferruginous Precambrian oceans (which clearly was not the case for these TOC-rich samples; Fig. 1), indicating extensive recycling of P back to the water column31,56. This group also includes ferruginous samples just above the Rooihoogte-Timeball Hill Formation boundary, where active microbial sulphate reduction is indicated by elevated Fepy/FeHR ratios (Figure S3), but not to the extent where euxinic bottom-waters developed (Fig. 1). By contrast, oxic, dysoxic and ferruginous samples deposited under probable lower seawater sulphate conditions commonly have Corg/Porg ratios closer to the Redfield ratio, with Corg/Preac ratios that fall below this value (Fig. 2). This is consistent with minimal preferential release of P from buried organic matter during early diagenesis (consistent with relatively low Fepy/FeHR ratios for samples above the Rooihoogte-Timeball Hill formation boundary; Figure S2), coupled with enhanced drawdown of P in association with the precipitation of Fe minerals25,28.

Climate and nutrient controls on oxygenation

The redox and P dynamics described above allow a timeline of environmental change, and its controls, to be reconstructed in the context of Snowball Earth deglaciation and the evolution of the GOE. We suggest, based on our data and others, an enhanced, rapid influx of P during Snowball Earth deglaciation, similar to that proposed for Neoproterozoic Snowball Earth events57–60. The consequent increase in productivity and organic carbon burial initially drove atmospheric and oceanic oxygenation during deposition of the middle Rooihoogte Formation (interval B; Fig. 1), but the oxygenation was transient (see below for further discussion). Nevertheless, this oxygenation promoted oxidative weathering on the continents and an enhanced influx of sulphate (indicated by low pyrite δ34S values), which led to the development of water column euxinia during deposition of interval C (Fig. 3A), and consequently, intense P recycling back to the water column (as indicated by high Corg/Porg and Corg/Preac ratios; Fig. 1).

Fig. 3. Illustrative model demonstrating the self-limiting feedbacks among the oxygen, sulphur and phosphorus cycles following glaciation.

Fig. 3

Open in a new tab

Panel progression from (A–C) denotes time since deglaciation, where (A) represents the post-glacial pulse of both phosphate and sulphate, which drove the initial rise of O2 through phosphorus recycling (intervals (A–C) in the text). B Demonstrates how feedbacks between phosphorus and sulphur led to progressive P exhaustion as euxinic conditions waned (interval (D) in the text). C Represents the new redox regime developed during deposition of the Timeball Hill Formation, whereby weathering fluxes returned to a new baseline and atmospheric oxygen level dropped from the peak values that occurred in the aftermath of deglaciation (interval (E) in the text).

The development of water column euxinia ultimately resulted in sulphate consumption via microbial sulphate reduction, resulting in lower seawater sulphate concentrations, as indicated by the progressive shift to higher δ34S values and a transition back to ferruginous water-column conditions in the lower part of interval D (Fig. 1). However, P recycling initially persisted (Figs. 1, 2) due to sufficient sulphate availability to drive pore-waters sulphidic, as indicated by moderate Fepy/FeHR ratios (Fig. S3). Ultimately, the positive productivity feedback imposed by P recycling across this interval further enhanced organic carbon production and burial (indicated by elevated Corg contents; Fig. 1), leading to a return to an oxygenated atmosphere across the Rooihoogte-Timeball Hill Formation boundary8,10. Furthermore, coupled to a probable progressive decrease in the dissolved P influx from continental weathering in the aftermath of glacial melting, the return of ferruginous seawater conditions ultimately resulted in the re-establishment of the PFe trap (Fig. 3B), as indicated by the subsequent transition back to low Corg/Preac ratios (Fig. 1). This negative nutrient feedback imposed by P drawdown in association with Fe minerals resulted in decreased water-column oxygen demand, through lower organic carbon production, and hence better oxygenated shallow waters.

Following these post-glacial dynamics, under oxic, dysoxic and ferruginous conditions through the remainder of intervals D and E, P recycling was limited or absent, despite occasionally elevated Corg/Porg and P/Al ratios (Fig. 1). These elevated Corg/Porg and P/Al ratios likely reflect enhanced P drawdown in association with Fe minerals under transient ferruginous conditions (as indicated by relatively high PFe contents for these samples; Table S4). In all cases, however, low Corg/Preac ratios support efficient sequestration of P in the sediment, either via the formation of authigenic minerals such as CFA33 or via trapping of upward-migrating P by iron minerals61. This sequestration of P would have restricted oxygen production, leading to atmospheric oxygen concentrations that were likely persistently low (potentially straddling the 10−6 PAL threshold). However, within this framework, a balance between low rates of productivity-driven oxygen production and consequent oxidation of atmospheric methane may have driven a progressive descent into the final, regional Paleoproterozoic glaciation1,42. This sequence of events is remarkably similar to those that occurred in the aftermath of the Neoproterozoic Sturtian glaciation52. Following the initial Sturtian deglaciation, a shift to euxinic conditions was followed by a transition to oxygenated conditions. For the majority of the studied locations, there is also a pronounced shift to highly elevated C:P ratios for both organic and reactive P, followed by a return to lower values52.

Constraining oxygen fluctuations during the GOE

To further evaluate the proposed feedbacks described above, we utilise an adapted biogeochemical model framework62 for the C, O, P and S cycles (Fig. 4; see Methods). The model was setup for generalised pre-GOE conditions through setting riverine phosphate inputs, and mantle and metamorphic reductant inputs, within established bounds to maintain a functionally anoxic atmosphere25. Phosphorus recycling rates in the model are assumed to be determined by a combination of both the pyrite burial rate and bottom-water oxygen content, reflecting the overall availability of sulphide in the water-column and pore-waters (see Methods). A post-glacial P and S weathering influx was approximated by back-calculating the original (prior to any recycling) P content of the sediment using the Redfield ratio and organic C contents, with the newly approximated P/Al ratios being compared to the average shale value63 (see Methods).

Fig. 4. Biogeochemical modelling results.

Fig. 4

Open in a new tab

The Δ33S and δ34S records are compared to the modelled atmospheric oxygen and sulphur reservoir sizes. Note that the modelled atmospheric O2 and sulphur reservoir scales are reversed to aid comparison with the geochemical data. Organic carbon (Corg) and total phosphorus (P) contents are compared to the organic carbon and phosphorus burial rates on the simulated continental shelf. Corg:Porg and Corg:Preac ratios are compared directly to modelled C:P ratios, with both the proximal and distal-shelf environment values shown. Corg – organic carbon, Porg – organic bound phosphorus, Preac - reactive phosphorus. R.- Rooihoogte Formation.

The model results indicate that the proposed post-glacial pulse of P and sulphate would have led to a substantial increase in productivity and atmospheric oxygen concentration during deposition of the Rooihoogte Formation (Fig. 4). However, as discussed above, an initial interval of atmospheric oxygenation was apparently interrupted by a transient return to anoxia in the upper Rooihoogte Formation (Fig. 1). This may be explained by a delayed pulse of degassing associated with the depressurisation of magmatic chambers following deglaciation64,65. While there is uncertainty in this depressurisation mechanism even for the Pleistocene glaciations, particularly with regard to the overall magnitude and timing of degassing66, our model results suggest that a delayed pulse of reduced gases would have been able to briefly return the atmosphere to anoxia, and when combined with the post-glacial weathering input, produces model outputs that are in close agreement with the geochemical data (Fig. 4). Specifically, the modelled atmospheric O2 and S reservoir dynamics match the sulphur isotope (Δ33S and δ34S) records reasonably well, while the distal shelf Corg, P and C:P ratios also respond in a similar manner to the data, with a rapid rise in Corg and C:P ratios associated with increased pyrite burial during the second oxygenation interval.

Importantly, having established the efficacy of the model, we are able to place constraints on the magnitude of O2 fluctuations and their controls (Fig. 4). Here, the initial rise in atmospheric oxygen in the middle of the Rooihoogte Formation, prior to delayed magmatic degassing, was driven by the increased P input, productivity and carbon burial following glacial melting, and this may have led to a relatively minor increase in atmospheric O2 concentration ( > 10−6 PAL). By contrast, the additional contribution to productivity and carbon burial from P recycling during the subsequent canonical loss of MIF-S at the height of post-glacial weathering (i.e., across the Rooihoogte-Timeball Hill Formation boundary) likely led to a major rise in atmospheric O2, with concentrations potentially approaching 10−2 PAL.

While we do not directly consider more complex atmospheric oxygen and methane dynamics during deposition of the overlying section, our reconstruction of P cycle dynamics highlights that a major drop in atmospheric O2 concentration would be expected1, due to the negative productivity feedback initiated by P sequestration into the sediments with the development of ferruginous conditions (Fig. 3B). Indeed, the full suite of data presented here suggests that minor changes in the balance between the fluxes of reductants and oxygen production would have been capable of readily driving variability in atmospheric oxygen levels at concentrations that may have straddled the 10−6 PAL MIF-S to MDF-S threshold. This possibility is supported by atmospheric modelling results, which have demonstrated that stable oxic and anoxic atmospheric states were both achievable, with rapid transitions between these states being dependent on the redox composition and fluxes of gases with respect to reservoir sizes67,68.

Our approach documents the impact of deglaciation on the behaviour of the P cycle within the context of the oscillatory rise of oxygen across the GOE. Glacial meltwaters brought appreciable bioavailable phosphate from the continents, ultimately driving an initial moderate increase in atmospheric O2 concentration. This led to enhanced oxidative weathering of the continents and hence an increased influx of sulphate, leading to the development of localised water-column euxinia and extensive P recycling. These feedbacks led to a major rise in atmospheric O2, however, as the rate of pyrite burial exceeded the rate of sulphate supply from oxidative weathering, the ocean transitioned back to a ferruginous state. This ultimately resulted in P drawdown, which, alongside a waning post-glacial influx of P, led to decreased net oxygen production and a newly established quasi-stable state. Subsequently, prior to the final, regional glaciation42, oxygen concentrations were maintained at an overall low level, but with oscillations potentially being driven by a changing balance between reductant inputs and oxygen production. The consistent, dynamic biogeochemical behaviour we observe between the phosphorus, carbon, sulphur and oxygen cycles suggests that the extreme climatic fluctuations that were a defining characteristic of the GOE drove dramatic variability in atmospheric and seawater oxygenation levels, with this instability persisting for over 200 million years.

Methods

Geochemical analyses

Samples were collected from the well-preserved EBA-2 drill-core at the National Core Library in Donkerhoek, South Africa (see Supplementary Information for a full description of the geological setting). Sulphur isotope and Corg concentration data were previously published1. Redox-sensitive trace elements (U, Mo, Re) and major elements (P and Al) were determined on ashed samples (450 °C overnight), followed by a HNO3–HF–HClO4–H3BO3 digest (full details, including precision and accuracy, are reported in the Supplementary Information).

Phosphorus phase partitioning was conducted following a revised extraction procedure for ancient rocks69. Four operationally-defined P phases were targeted, including iron-bound P (PFe), authigenic P (Paut), organic-bound P (Porg) and crystalline apatite (predominantly detrital P; Pcryst). Phosphorus concentrations were either determined spectrophotometrically using the molybdate-blue method at 880 nm70, or via ICP-OES for extracts containing chemical compounds that interfere with colour development via the molybdate-blue method69. Reactive P (Preac) is quantified as the sum of PFe, Paut and Porg31. Full details, including precision and accuracy, and consideration of the potential impact of transfer of Preac to Pcryst, are reported in the Supplementary Information.

Biogeochemical Modelling

We utilize a 5-box ocean-atmosphere biogeochemical model for the C-O-P cycles (modified from ref. 62), supplemented by a 1-box marine sulphur cycle, to examine the response of the phosphorus cycle to a changing seawater sulphate reservoir and sulphur fluxes. This S cycle framework is based upon the single-box COPSE model framework71. Both frameworks are briefly summarised here, with a breakdown of the model fluxes and reservoir calculations provided in Tables S1–3. The model consists of four ocean boxes62, representing the proximal and distal shelf along with the surface and deep oceans, and within them, the organic C, organic P, soluble P and oxygen cycles are tracked. An atmosphere box is also included, which only considers oxygen. The P cycle includes the three major burial phases of P, namely organic-bound P (Porg), iron-oxide-bound P (PFe) and authigenic P (Paut).

The addition of the S cycle includes a single box encompassing all regions of the ocean, where reduced and oxidized forms of S are included. Inputs of S include pyrite weathering, which is calculated based on a previously published oxidative weathering relationship (72; Eqn. S76). This assumption allows for all pyrite to be oxidized when pO2 is greater than 0.35% PAL (transport-limited regime), with almost all pyrite remaining unoxidized at pO2 levels lower than ~3 × 10−4% PAL (reaction-limited)72. We also consider the assumption73 that only 60% of continental pyrite could be weathered due to the absence of land plants. Reduced sulphur degassing is also considered in a similar formulation to COPSE71, but without the assumption of a changing crustal pyrite inventory, given the relatively short run-time of this model (Eqn. S78). Pyrite weathering is considered a sink of oxygen in the same manner as the oxidative weathering of organic matter. Gypsum weathering and oxidised sulphur degassing are not considered here, as the earliest reported appreciable sulphate evaporite deposits are at ~2.4 Ga16,17,74, meaning that if gypsum was extensively buried at this time, it would likely not have been recycled through subduction and uplift in time to be weathered and metamorphically degassed during deposition of the studied interval. Burial of reduced S as pyrite is calculated based on both organic C availability (organic C burial flux) and the S reservoir size, following COPSE (Eqn. S80). This pyrite burial flux is then introduced as an oxygen source to the atmosphere71. A reduced gas input appropriate for pre-GOE conditions is imposed on the model, as in previous work25,62.

A pulse of both P and sulphate through post-glacial weathering is introduced under the assumption that this immediately followed the Snowball Earth glaciation. This input flux is calculated based on the initial burial proportion of organic C and P in order to back-calculate a P/Al ratio of the sediments prior to diagenetic recycling. These ratios are then compared to the Phanerozoic average shale P/Al ratio in order to provide average and maximum changes relative to the present day. By distinguishing between oxic and euxinic samples, minimal and maximum input fluxes can be estimated. This amounts to 70% of the present-day P input prior to and following deglaciation, and values ranging from 196 to 733% of the present-day input represent the post-glacial pulse. Based on this range, and informed by our dataset, we use a conservative input of 170%. This flux was then applied over a period within estimates of Snowball Earth deglaciation rates ( ~ 100 kyrs75). In addition to this, we include the impact of depressurized magma chambers during large-scale deglaciation, leading to an excess pulse of magmatic gases64. This was included as an additional input of reduced gases to the atmosphere (10 × 1012) and aids in reproducing the initial rapid loss and subsequent return of the MIF-S signal during deposition of the Rooihoogte Formation.

Supplementary information

Supplementary Information (1.1MB, docx)
Transparent Peer Review file (348.1KB, pdf)

Author contributions

S.W.P., L.J.A., B.J.W.M., and A.B. designed the research. S.W.P and A.B. collected the samples. L.J.A. and Z.P. performed geochemical analyses. L.J.A. and B.J.W.M. conducted the biogeochemical modeling, and A.B. provided the geological context. All authors contributed to writing the manuscript.

Peer review

Peer review information

Nature Communications thanks Wei Wei and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

L.J.A. was funded by a Leeds Anniversary Research Scholarship. We acknowledge financial support from NERC grants NE/R010129/1 (S.W.P., B.J.W.M.), and NE/T008458/1 (S.W.P.). Participation of A.B. and Z.P. was made possible with funding from NSERC Discovery and Accelerator grants and the China Scholarship Council (CSC NO. 202004910064), respectively.

Data availability

The data generated in this study are provided in the Supplementary Information and is available in the Figshare repository https://doi.org/10.6084/m9.figshare.3289559376.

Code availability

Code is available at https://github.com/lalcott/oneboxScycle.git and https://doi.org/10.5281/zenodo.2116216977.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76597-y.

References

  • 1.Poulton, S. W. et al. A 200-million-year delay in permanent atmospheric oxygenation. Nature592, 232–236 (2021). [DOI] [PubMed] [Google Scholar]
  • 2.Poulton, S. W. & Canfield, D. E. Ferruginous conditions: a dominant feature of the ocean through Earth’s history. Elements7, 107–112 (2011). [Google Scholar]
  • 3.Lyons, T. W., Reinhard, C. T. & Planavsky, N. J. The rise of oxygen in Earth’s early ocean and atmosphere. Nature506, 307–315 (2014). [DOI] [PubMed] [Google Scholar]
  • 4.Canfield, D. E. The early history of atmospheric oxygen: homage to Robert M. Garrels. Annu Rev. Earth Planet Sci.33, 1–36 (2005). [Google Scholar]
  • 5.Farquhar, J., Bao, H. & Thiemens, M. Atmospheric influence of Earth’s earliest sulfur cycle. Science289, 756–758 (2000). [DOI] [PubMed] [Google Scholar]
  • 6.Gumsley, A. P. et al. Timing and tempo of the great oxidation event. PNAS114, 1811–1816 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Warke, M. R. et al. The great oxidation event preceded a paleoproterozoic “snowball Earth. PNAS117, 13314–13320 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Bekker, A. et al. Dating the rise of atmospheric oxygen. Nature427, 117–120 (2004). [DOI] [PubMed] [Google Scholar]
  • 9.Guo, Q. et al. Reconstructing Earth’s surface oxidation across the Archean-Proterozoic transition. Geology37, 399–402 (2009). [Google Scholar]
  • 10.Luo, G. et al. Rapid oxygenation of Earth’s atmosphere 2.33 billion years ago. Sci. Adv.2, e1600134 (2016). [DOI] [PMC free article] [PubMed]
  • 11.Izon, G. et al. Bulk and grain-scale minor sulfur isotope data reveal complexities in the dynamics of Earth’s oxygenation. PNAS119, e2025606119 (2022). [DOI] [PMC free article] [PubMed]
  • 12.Uveges, B. T., Izon, G., Ono, S., Beukes, N. J. & Summons, R. E. Reconciling discrepant minor sulfur isotope records of the Great Oxidation Event. Nat. Commun.14, 279 (2023). [DOI] [PMC free article] [PubMed]
  • 13.Zahnle, K. J., Claire, M. & Catling, D. C. The loss of mass-independent fractionation in sulfur due to a Palaeoproterozoic collapse of atmospheric methane. Geobiology4, 271–283 (2006). [Google Scholar]
  • 14.Catling, D. C. & Zahnle, K. J. The Archean atmosphere. Sci. Adv.6, 10.1126/sciadv.aax1420 (2020). [DOI] [PMC free article] [PubMed]
  • 15.Bekker, A. & Kaufman, A. J. Oxidative forcing of global climate change: a biogeochemical record across the oldest Paleoproterozoic ice age in North America. Earth Planet. Sci. Lett.258, 486–499 (2007). [Google Scholar]
  • 16.Rasmussen, B., Bekker, A. & Fletcher, I. R. Correlation of Paleoproterozoic glaciations based on U-Pb zircon ages for tuff beds in the Transvaal and Huronian Supergroups. Earth Planet. Sci. Lett.382, 173–180 (2013). [Google Scholar]
  • 17.Rasmussen, B., Zi, J. & Bekker, A. New U-Pb zircon tuff ages and revised stratigraphic correlations in the Superior craton during the Great Oxidation Episode. Earth Planet. Sci. Lett.640, 118779 (2024).
  • 18.Hoffman, P. F. The Great Oxidation and a Siderian snowball Earth: MIF-S based correlation of Paleoproterozoic glacial epochs. Chem. Geol.362, 143–156 (2013). [Google Scholar]
  • 19.Garduno Ruiz, D., Goldblatt, C. & Ahm, A.-S. Climate variability leads to multiple oxygenation episodes across the Great Oxidation Event. Geophys. Res. Lett.51, e2023GL106694 (2024). [Google Scholar]
  • 20.Garduno Ruiz, D., Goldblatt, C. & Ahm, A.-S. Climate shapes the oxygenation of Earth’s atmosphere across the Great Oxidation Event. Earth Planet. Sci. Lett.607, 118071 (2023). [Google Scholar]
  • 21.Reinhard, C. T., Planavsky, N. J. & Lyons, T. W. Long-term sedimentary recycling of rare sulfur isotope anomalies. Nature497, 100–103 (2013). [DOI] [PubMed] [Google Scholar]
  • 22.Killingsworth, B. A. et al. Constraining the rise of oxygen with oxygen isotopes. Nat. Commun.10, 4924 (2019). [DOI] [PMC free article] [PubMed]
  • 23.Ostrander, C. M. et al. Onset of coupled atmosphere–ocean oxygenation 2.3 billion years ago. Nature631, 335–339 (2024). [DOI] [PubMed] [Google Scholar]
  • 24.Tyrrell, T. The relative influences of nitrogen and phosphorus on oceanic primary production. Nature400, 525–531 (1999). [Google Scholar]
  • 25.Alcott, L. J., Mills, B. J. W., Bekker, A. & Poulton, S. W. Earth’s Great Oxidation Event facilitated by the rise of sedimentary phosphorus recycling. Nat. Geosci.15, 210–215 (2022). [Google Scholar]
  • 26.Cox, G. M., Lyons, T. W., Mitchell, R. N., Hasterok, D. & Gard, M. Linking the rise of atmospheric oxygen to growth in the continental phosphorus inventory. Earth Planet. Sci. Lett.489, 28–36 (2018). [Google Scholar]
  • 27.Reinhard, C. T. et al. Evolution of the global phosphorus cycle. Nature541, 386–389 (2017). [DOI] [PubMed] [Google Scholar]
  • 28.Guilbaud, R. et al. Phosphorus-limited conditions in the early Neoproterozoic ocean maintained low levels of atmospheric oxygen. Nat. Geosci.13, 296–301 (2020). [Google Scholar]
  • 29.Krom, M. D. & Berner, A. B. R. The diagenesis of phosphorus in a nearshore marine sediment. Geochimica et. Cosmochinica Acta45, 207–216 (1981). [Google Scholar]
  • 30.Slomp, C. P., Thomson, J. & de Lange, G. J. Enhanced regeneration of phosphorus during formation of the most recent eastern Mediterranean sapropel (S1). Geochimica et. Cosmochinica Acta66, 1171–1184 (2002). [Google Scholar]
  • 31.Ingall, E. & Jahnke, R. Evidence for enhanced phosphorus regeneration from marine sediments overlain by oxygen-depleted waters. Geochimica et. Cosmochinica Acta58, 2571–2575 (1994). [Google Scholar]
  • 32.Ingall, E. D., Bustin, R. M. & Van Cappellen, P. Influence of water column anoxia on the burial and preservation of carbon and phosphorus in marine shales. Geochimica et. Cosmochimica Acta57, 303–316 (1993). [Google Scholar]
  • 33.Ruttenberg, K. C. & Berner, R. A. Authigenic apatite formation and burial in sediments from non-upwelling, continental margin environments. Geochimica et. Cosmochimica Acta57, 991–1007 (1993). [Google Scholar]
  • 34.März, C., Riedinger, N., Sena, C. & Kasten, S. Phosphorus dynamics around the sulfate-methane transition in continental margin sediments: authigenic apatite and Fe(II) phosphates. Mar. Geol.404, 84–96 (2018). [Google Scholar]
  • 35.Slomp, C. P. et al. Coupled dynamics of iron and phosphorus in sediments of an oligotrophic coastal basin and the impact of anaerobic oxidation of methane. PLoS One8, 62386 (2013). [DOI] [PMC free article] [PubMed]
  • 36.Xiong, Y. et al. Phosphorus cycling in Lake Cadagno, Switzerland: a low sulfate euxinic ocean analogue. Geochimica et. Cosmochimica Acta251, 116–135 (2019). [Google Scholar]
  • 37.Van Cappellen, P. & Ingall, E. D. Benthic phosphorus regeneration, net primary production, and ocean anoxia: a model of the coupled marine biogeochemical cycles of carbon and phosphorus. Paleoceanography9, 677–692 (1994). [Google Scholar]
  • 38.Zerkle, A. L. et al. Onset of the aerobic nitrogen cycle during the Great Oxidation Event. Nature542, 465–467 (2017). [DOI] [PubMed] [Google Scholar]
  • 39.Li, S., Wignall, P. B. & Poulton, S. W. Co-application of rhenium, vanadium, uranium and molybdenum as paleo-redox proxies: insight from modern and ancient environments. Chem. Geol.674, 122565 (2025).
  • 40.Eriksson, K. A. The Timeball Hill Formation - a fossil delta. J. Sed. Petrol.43, 1046–1053 (1973). [Google Scholar]
  • 41.Visser, J. N. J. The Timeball Hill Formation at Pretoria — a prograding shoreline deposit. Ann. Geol. Surv. S. Afr.9, 115–118 (1972). [Google Scholar]
  • 42.Bekker, A., Krapež, B. & Karhu, J. A. Correlation of the stratigraphic cover of the Pilbara and Kaapvaal cratons recording the lead up to Paleoproterozoic Icehouse and the GOE. Earth-Sci. Rev.211, 103389 (2020). [Google Scholar]
  • 43.Anderson, R. F., Fleisher, M. Q. & LeHuray, A. P. Concentration, oxidation state, and particulate flux of uranium in the Black Sea. Geochimica et. Cosmochimica Acta53, 2215–2224 (1989). [Google Scholar]
  • 44.Crusius, J., Calvert, S., Pedersen, T. & Sage, D. Rhenium and molybdenum enrichments in sediments as indicators of oxic, suboxic and sulfidic conditions of deposition. Earth Planet. Sci. Lett.145, 65–78 (1996). [Google Scholar]
  • 45.Morford, J. L. & Emerson, S. The geochemistry of redox sensitive trace metals in sediments. Geochimica et. Cosmochimica Acta63, 1735–1750 (1999). [Google Scholar]
  • 46.Erickson, B. E. & Helz, G. R. Molybdenum(VI) speciation in sulfidic waters: stability and lability of thiomolybdates. Geochimica et. Cosmochimica Acta64, 1149–1158 (2000). [Google Scholar]
  • 47.Helz, G. R. et al. Mechanism of molybdenum removal from the sea and its concentration in black shales: EXAFS evidence. Geochimica et. Cosmochimica Acta60, 3631–3642 (1996). [Google Scholar]
  • 48.Tribovillard, N., Algeo, T. J., Baudin, F. & Riboulleau, A. Analysis of marine environmental conditions based on molybdenum-uranium covariation—Applications to Mesozoic paleoceanography. Chem. Geol.324-325, 46–58 (2012). [Google Scholar]
  • 49.McLennan, S. M. Relationships between the trace element composition of sedimentary rocks and upper continental crust. Geochem. Geophys. Geosyst.2, 109 (2001).
  • 50.Poulton, S. W. Early phosphorus redigested. Nat. Geosci.10, 75–76 (2017). [Google Scholar]
  • 51.Halevy, I. et al. Sedimentary parameters control the sulfur isotope composition of marine pyrite. Science382, 946–951 (2023). [DOI] [PubMed] [Google Scholar]
  • 52.Pasquier, V., Fike, D. A., Révillon, S. & Halevy, I. A global reassessment of the controls on iron speciation in modern sediments and sedimentary rocks: a dominant role for diagenesis. Geochimica et. Cosmochimica Acta335, 221–230 (2022). [Google Scholar]
  • 53.Habicht, K. S., Gade, M., Thamdrup, B., Berg, P. & Canfield, D. E. Calibration of sulfate levels in the Archean Ocean. Science298, 2372–2374 (2002). [DOI] [PubMed] [Google Scholar]
  • 54.Pasquier, V., Bryant, R. N., Fike, D. A. & Halevy, I. Strong local, not global, controls on marine pyrite sulfur isotopes. Sci. Adv.7, eabb7403 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Canfield, D. E. & Raiswell, R. The evolution of the sulfur cycle. Am. J. Sci.299, 697–723 (1999). [Google Scholar]
  • 56.Ingall, E. & Jahnke, R. Influence of water-column anoxia on the elemental fractionation of carbon and phosphorus during sediment diagenesis. Mar. Geol.139, 219–229 (1997). [Google Scholar]
  • 57.Bowyer, F. et al. Biological diversification linked to environmental stabilization following the Sturtian Snowball glaciation. Sci. Adv.9, 9999 (2023). [DOI] [PMC free article] [PubMed]
  • 58.Planavsky, N. et al. The evolution of the marine phosphate reservoir. Nature467, 1088–1090 (2010). [DOI] [PubMed] [Google Scholar]
  • 59.Harada, M., Tajika, E. & Sekine, Y. Transition to an oxygen-rich atmosphere with an extensive overshoot triggered by the Paleoproterozoic snowball Earth. Earth Planet. Sci. Lett.419, 178–186 (2015). [Google Scholar]
  • 60.Harada, M., Miura, Y., Watanabe, Y. & Ozaki, K. Redox dynamics of the atmosphere and oceans induced by the paleoproterozoic Snowball Earth Events. Geobiology23, e70040 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Slomp, C. P., Epping, E. H. G., Helder, W. & Van Raaphorst, W. A key role for iron-bound phosphorus in authigenic apatite formation in North Atlantic continental platform sediments. J. Mar. Res.54, 1179–1205 (1996). [Google Scholar]
  • 62.Alcott, L. J., Mills, B. J. W. & Poulton, S. W. Stepwise Earth oxygenation is an inherent property of global biogeochemical cycling. Science366, 1333–1337 (2019). [DOI] [PubMed] [Google Scholar]
  • 63.Turekian, K. K. & Wedepohl, K. H. Distribution of the elements in some major units of the Earth’s crust. GSA Bull.72, 175–192 (1961). [Google Scholar]
  • 64.Huybers, P. & Langmuir, C. Feedback between deglaciation, volcanism, and atmospheric CO2. Earth Planet. Sci. Lett.286, 479–491 (2009). [Google Scholar]
  • 65.Hansenclever, J. et al. Sea level fall during glaciation stabilized atmospheric CO2 by enhanced volcanic degassing. Nat. Commun.8, 15867 (2017). [DOI] [PMC free article] [PubMed]
  • 66.Ganopolski, A. & Brovkin, V. Simulation of climate, ice sheets and CO2 evolution during the last four glacial cycles with an Earth system model of intermediate complexity. Climate13, 1695–1716 (2017). [Google Scholar]
  • 67.Gregory, B. S., Claire, M. W. & Rugheimer, S. Photochemical modelling of atmospheric oxygen levels confirms two stable states. Earth Planet. Sci. Lett.561, 116818 (2021). [Google Scholar]
  • 68.Wogan, N. F., Catling, D. C., Zahnle, K. J. & Claire, M. W. Rapid timescale for an oxic transition during the Great Oxidation Event and the instability of low atmospheric O2. PNAS119, e2205618119 (2022). [DOI] [PMC free article] [PubMed]
  • 69.Thompson, J. et al. Development of a modified SEDEX phosphorus speciation method for ancient rocks and modern iron-rich sediments. Chem. Geol.524, 383–393 (2019). [Google Scholar]
  • 70.Strickland, J. D. H. & Parsons, T. R. A practical handbook of seawater analysis. Fish. Res. Board Can. Ott. Bull.167, 45–64 (1972). [Google Scholar]
  • 71.Tostevin, R. & Mills, B. J. W. Reconciling proxy records and models of Earth’s oxygenation during the Neoproterozoic and Palaeozoic. Interface Foxus10, 20190137 (2020). [DOI] [PMC free article] [PubMed]
  • 72.Ozaki, K., Reinhard, C. T. & Tajika, E. A sluggish mid-Proterozoic biosphere and its effect on Earth’s redox balance. Geobiology17, 3–11 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 73.Stüeken, E. E., Catling, D. C. & Buick, R. Contributions to late Archaean sulphur cycling by life on land. Nat. Geosci.5, 722–725 (2012). [Google Scholar]
  • 74.Yudoskaya, M. A. et al. in Metals, Minerals, and Society Vol. 21 (eds A. M Arribas & J. L Mauk) (Society of Economic Geologists, 2018).
  • 75.Crowley, T. J., Hyde, W. T. & Peltier, W. R. CO2 levels required for deglaciation of a “near-snowball” Earth. Geophys. Res. Lett.28, 283–286 (2001). [Google Scholar]
  • 76.Alcott, L. J., Mills, B. J. W., Bekker, A., Peng, Z. & Poulton, S. W. A nutrient control on oxygenation dynamics during Earth’s Great Oxidation Episode, version 1, Figshare 10.6084/m9.figshare.32895593. (2026). [DOI] [PMC free article] [PubMed]
  • 77.Alcott, L. J., Mills, B. J. W., Bekker, A., Peng, Z. & Poulton, S. W. A nutrient control on oxygenation dynamics during Earth’s Great Oxidation Episode, version 1, Zenodo 10.5281/zenodo.21162169. (2026). [DOI] [PMC free article] [PubMed]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information (1.1MB, docx)
Transparent Peer Review file (348.1KB, pdf)

Data Availability Statement

The data generated in this study are provided in the Supplementary Information and is available in the Figshare repository https://doi.org/10.6084/m9.figshare.3289559376.

Code is available at https://github.com/lalcott/oneboxScycle.git and https://doi.org/10.5281/zenodo.2116216977.

Articles from Nature Communications are provided here courtesy of Nature Publishing Group

RESOURCES