Earth and Planetary Science Letters Moderate levels of oxygenation during the late stage of Earth’s Great Oxidation Event

represents Earth’s most pronounced and longest-lived positive carbon isotope excursion. However, the magnitude and extent of atmosphere-ocean oxygenation and implications for the biosphere during this critical period in Earth’s history remain poorly constrained. Here, we present nitrogen (N), selenium (Se), and carbon (C) isotope data, as well as bio-essential element concentrations, for Paleoproterozoic marine shales deposited during the LE. The data provide evidence for a highly productive and well-oxygenated photic zone, with both inner and outer-shelf marine environments characterized by nitrate- and Se oxyanion-replete conditions. However, the redoxcline subsequently encroached back onto the inner shelf during global-scale deoxygenation of the atmosphere-ocean system at the end of the LE, leading to locally enhanced water column denitriﬁcation and quantitative reduction of selenium oxyanions. We propose that nitrate-replete conditions associated with fully oxygenated continental shelf settings were a common feature during the LE, but nitriﬁcation was not suﬃciently widespread for the aerobic nitrogen cycle to impact the isotopic composition of the global ocean N inventory. Placed in the context of Earth’s broader oxygenation history, our ﬁndings indicate that O 2 levels in the atmosphere-ocean system were likely much lower than modern concentrations. Early Paleoproterozoic biogeochemical cycles were thus far less advanced than after Neoproterozoic oxygenation. © 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Biological evolution and ecosystem complexity are intimately linked to the magnitude and expansion of Earth's surface oxygenation through time. Constraining the timing at which the atmosphere-hydrosphere system became significantly oxygenated, reaching near-modern levels, may help to shed new light on the portant markers for the GOE (Bekker et al., 2004;Farquhar et al., 2000;Holland, 2006;Lyons et al., 2014). However, the permanent rise of atmospheric O 2 likely occurred in the second half of the GOE, during the ∼2.22-2.06 Ga Lomagundi Event (LE) (Poulton et al., 2021). The LE represents the most pronounced and longestlived positive carbon isotope excursion in Earth history (Bekker, 2022;Karhu and Holland, 1996), and is considered to have resulted in transient highly elevated levels of O 2 , but to an extent that is currently unclear (Bekker and Holland, 2012). Atmospheric O 2 concentrations are estimated to have been at levels below 1% (Colwyn et al., 2019) to as high as ∼10% PAL (Holland, 2006;Kanzaki and Murakami, 2016) throughout most of the subsequent Proterozoic, up until the ∼750-539 Ma Neoproterozoic Oxygenation Event (NOE), when atmospheric O 2 is commonly considered to have increased further (Holland, 2006;Lyons et al., 2014;Colwyn et al., 2019).
Mainly based on carbon isotope systematics, the LE is generally considered to reflect a combination of high rates of primary productivity and increased organic carbon burial (Bekker, 2022;Bekker and Holland, 2012;Husson and Peters, 2017;Karhu and Holland, 1996). Elevated organic productivity may have been initiated by a high phosphorus flux to the oceans, potentially due to leaching of terrestrial rocks by sulfuric acid generated through oxidative weathering of pyrite (Bekker and Holland, 2012;Konhauser et al., 2011). The rate of organic matter burial exceeded the rate of its reoxidation, allowing the release of marine photosynthetic O 2 which likely caused an atmospheric O 2 overshoot (Bekker and Holland, 2012). However, redox and biogeochemical proxies used for constraining O 2 concentrations in the atmosphere-ocean system during the LE have yielded contrasting estimates (Blättler et al., 2018;Colwyn et al., 2019;Hardisty et al., 2014;Holland, 2006;Kanzaki and Murakami, 2016;Kipp et al., 2017;Mänd et al., 2020). For example, estimates based on the chemical and Cr isotope compositions of Paleoproterozoic paleosols suggest that atmospheric O 2 levels remained lower than 1-10% PAL (Colwyn et al., 2019;Kanzaki and Murakami, 2016). By contrast, thick (∼800 m) evaporite deposits supposedly deposited in marginal-marine environments during the LE, which are interpreted to reflect the oceanic sulfate inventory, as well as associated S, Ca, Sr and O isotope compositions, have been proposed to reflect an oxygenated atmosphere-ocean system with higher than 20% of the modern atmosphere-ocean oxidizing capacity (Blättler et al., 2018).
Although the mechanism that caused the end of the LE is still debated, its termination, based on C, S, Fe, V, Mo, U and Se biogeochemical proxies in marine sediments, is viewed to be contemporaneous with a drop in O 2 levels in the atmosphere-ocean system (Asael et al., 2018;Canfield et al., 2013;Kipp et al., 2017;Kump et al., 2011;Lyons et al., 2014;Ossa Ossa et al., 2018, 2021aPlanavsky et al., 2012;Scott et al., 2014). However, based on U and Cr isotope compositions and trace-metal (Mo, Cr, U and Re) enrichments in mid-Paleoproterozoic shales from the Zaonega Formation, northwestern Russia, it has been proposed that the atmosphereocean system was still well-oxygenated millions of years after the termination of the LE (Mänd et al., 2020(Mänd et al., , 2022. Since the biogeo-chemical cycling of S, Fe, V, Mo and U yield stable oxyanions at circumneutral pH in marine environments at relatively low redox potential (i.e., Eh < 0 V; Rue et al., 1997;Stumm and Morgan, 1970), they are only sensitive as redox indicators at very low levels of environmental O 2 (e.g., Andersen et al., 2020;Dang et al., 2022). Furthermore, a wide range of redox and non-redox processes can significantly fractionate Cr isotopes, regardless of the level of O 2 available in the environment (e.g., Babechuk et al., 2017Babechuk et al., , 2018Miletto et al., 2021). Clearly, further constraints are needed given current uncertainties in estimates of environmental O 2 availability. Marine proxies that are redox-sensitive at a higher oxygen threshold (i.e., Eh 0 V) at circumneutral pH are thus required to shed light on the potential for higher oxygenation levels during the LE.
To provide new insight into the evolution of oceanic redox structure during and after the LE and links to environmental oxygenation, we present combined nitrogen and selenium isotope ratios, together with new Fe speciation and organic carbon isotope data, for well-preserved black shales from the mid-Paleoproterozoic Francevillian Group (Gauthier-Lafaye and Weber, 2003;Ossa Ossa et al., 2018, 2021a, which we compare with other LE sedimentary successions. These isotopic systems allow us to track the marine biogeochemical cycles of nitrogen and selenium in redox-stratified basins (Ader et al., 2016;Rue et al., 1997;Sigman and Fripiat, 2019;Stüeken, 2017;Stüeken et al., 2016), as well as the behavior of nitrate (NO 3 − ) and selenate (SeO 4 2− ), both of which are stable at much higher redox thresholds than SO 4 2− , MoO 4 2− , Fe 3+ and U 6+ (Rue et al., 1997;Stumm and Morgan, 1970). Both nitrate and selenate are stable at Eh > 0.4 V at circumneutral pH, and only build up to significant levels in openmarine environments if the water column is well oxygenated (Rue et al., 1997). Tracing the biogeochemical nitrogen cycle is complicated, because the nitrogen isotope composition of sedimentary rocks can be influenced by a wide range of both biotic and abiotic processes (Ader et al., 2016;Sigman and Fripiat, 2019;Stüeken et al., 2016). However, when paired with the Se biogeochemical cycle, a more rigorous constraint may be achieved, whereby the behavior of NO 3 − and SeO 4 2− is expected to be internally consistent under oxic, suboxic and anoxic conditions, and this consistency should be recorded in the isotopic signatures. Combined N and Se isotope analyses thus provide a powerful opportunity to unravel the magnitude and extent of ocean oxygenation during and shortly after the LE.

Major elements
Powdered samples were analysed for major element concentrations by X-ray fluorescence spectroscopy. Analyses were carried out on fusion beads, using a PANalytical MagiX Pro PW2540 spectrometer at the University of Johannesburg. Accuracy was checked with certified reference materials and was better than 1%. Elemental concentrations are reported in wt.% with a detection limit of 0.004 wt.%.

Iron speciation analysis
Iron speciation analyses was performed at the University of Leeds, UK using a calibrated sequential extraction protocol followed by Fe analysis via AAS (Poulton et al., 2021;Poulton and Canfield, 2005). This method is designed to operationally quantify four different pools of Fe considered to be highly-reactive (Fe HR ) towards sulfide in surface and near-surface environments: (a) pyrite S extracted via Cr-reduction followed by trapping as Ag 2 S, with Fe calculated assuming an FeS 2 stoichiometry (Fe Py ); (b) carbonate-associated iron extracted with a sodium acetate solution (Fe Carb ); (c) ferric oxides extracted with a dithionite solution (Fe Ox ); and (d) mixed-valence iron oxides, principally magnetite, extracted using ammonium oxalate (Fe Mag ). The total iron content in ancient marine shales (Fe T ) represents the sum of Fe HR and Fe bound in silicates (Poulton et al., 2021;Poulton and Canfield, 2005). Samples were run alongside an international Fe speciation standard (WHIT; Alcott et al., 2020), with replicate analyses giving a RSD of <5% for all Fe pools. For a detailed description of the interpretation of Fe speciation data, see Supplementary Information SI 2.

Carbon and nitrogen isotope analyses
The carbon and nitrogen stable isotope composition of bulk rock (δ 13 C org and δ 13 C bulk , respectively) and the nitrogen isotope composition of the separated kerogen (δ 15 N ker ) were determined by elemental analysis/isotope ratio mass spectrometry (EA/IRMS; Flash EA coupled to a Thermo Scientific Delta V isotope ratio mass spectrometer via a Conflow III interface) at the Institute of Earth Surface Dynamics, University of Lausanne (IDYST-UNIL). Kerogen, the fraction of organic matter insoluble in organic solvents, was separated from powdered shale samples using nonoxidizing acids (HCl, HF) to dissolve the mineral matrix. For δ 13 C and δ 15 N analyses, separate EA combustions were performed using sample aliquots with a 1:50 weight size difference. The isotope ratios are expressed in conventional delta (δ) notation as the per mil ( ) of 13 C/ 12 C and 15 N/ 14 N of the sample relative to the standard VPDB for δ 13 C and Air-N 2 for δ 15 N. The measured isotopic ratios were converted to the international scales with a 3-or 4-point calibration using international reference materials (IRMs; USGS-24, USGS-40, USGS-41, IAEA-600 or USGS64, USGS65, and USGS66) and inhouse standards. The accuracy and precision of the analyses were checked periodically through the analysis of standards not included as calibration standards. Separate aliquots of the USGS SGR-1b standard (petroleum and carbonate-rich shale) were decarbonated and analysed for δ 13 C and δ 15 N values. The obtained values (δ 13 C = −29.29 ± 0.17 , V-PDB; n = 10; δ 15 N = 17.45 ± 0.26 , Air-N 2 ; n = 6) are in good agreement with the values reported by Dennen et al. (2006) for SGR-1 (δ 13 C = −29.3 ± 0.3 , n = 27; δ 15 N = 17.4 ± 0.9 , n = 18). For a detailed description of the analytical procedures, see Supplementary Information SI 2.

Selenium isotope and elemental analyses
Sample preparation and analysis was carried out at the University of Tuebingen, Germany. Selenium isotope composition was determined by the double-spike method using a ThermoFisher Scientific NeptunePlus multicollector inductively coupled plasma mass spectrometer (MC-ICP-MS) coupled with an HGX-200 hydride generator (Kurzawa et al., 2017). Data are reported in δ 82/76 Se notation, that is, the per mil ( ) variation of 82 Se/ 76 Se relative to the NIST SRM 3149 standard. The accuracy and measurement precision were checked by analysing interlaboratory standard MH 495, international rock reference materials and for sulfide analyses in particular by repeated digestion of the Phanerozoic Navajún pyrite (König et al., 2019;Kurzawa et al., 2017). The MH 495 measured during the course of this study yielded a mean δ 82/76 Se value of −3.24 ± 0.05 (2 SD, n = 28), which is consistent with previously reported data, while the in-house reference Navajún pyrite (König et al., 2019 and references therein) yielded a mean δ 82/76 Se value of −2.45 ± 0.10 (2 SD, n = 4), which is within the external reproducibility as derived from multiple analytical sessions over 16 months, with a mean δ 82/76 Se value of −2.65 ± 0.20 (2 SD, n = 15). For a detailed description of the analytical procedures, see Supplementary Information SI 2.

Results
The description of the main results (Tables S1, S2) is organized with respect to previous characterization of redox conditions coupled to major upwelling and/or transgression events during deposition of the FB Formation and FC 1 Member (Figs. 1B, 2, S2B; Canfield et al., 2013;Ossa Ossa et al., 2018). For example, published iron speciation data based on highly reactive (Fe HR ) to total Fe (Fe T ) ratios indicate that deposition of the lower FB Formation occurred in a well-oxygenated marine environment (Fe HR /Fe T < 0.22), which subsequently evolved to a suboxic and even euxinic (anoxic and sulfidic; denoted by Fe HR /Fe T > 0.38 and pyrite Fe (Fe Py ) to Fe HR ratios > 0.7) water column setting during deposition of the upper part of the FB 1c unit (Figs. 2, S2B; Canfield et al., 2013;Ossa Ossa et al., 2013. Oxic conditions during deposition of most of the FB Formation, with suboxic conditions during deposition of the upper part of the FB 1c unit, are further supported by V/Al ratios averaging 7.7 ppm/wt.%, Mo/Al ratios averaging 0.2 ppm/wt.% and U/Al ratios averaging 7.7 ppm/wt.%, which are typical for marine black shale deposited under oxic conditions (Ossa Ossa et al., 2021a). By contrast, higher ratios of V/Al averaging 46.5 ppm/wt.%, Mo/Al averaging 1.6 ppm/wt.%, and U/Al averaging 1.2 ppm/wt.% in the Upper FC 1 Member and FD Formation indicate enhanced seawater euxinia during deposition of the upper part of the Francevillian Group (Canfield et al., 2013;Ossa Ossa et al., 2021a). This is further supported by Fe speciation data, with Fe HR /Fe T > 0.38 and Fe Py /Fe HR > 0.7 in the FD Formation (Canfield et al., 2013;Fig. S2).
In the lower part of the FB Formation ( Fig. 2; Table S2), black shales of the FB 1a unit, which record the main transgressive event in the basin (Canfield et al., 2013;Gauthier-Lafaye and Weber, 2003;Ossa Ossa et al., 2013;Weber, 1968), are characterized by Se concentrations between 2 and 6 μg/g, with slightly negative δ 82 Se values between −0.35 and −0.28 (mean value of −0.32 ). The δ 15 N values of bulk samples (δ 15 N bulk ) and extracted kerogen (δ 15 N ker ) show no systematic offset and are positive, with values between +4.4 and +8.1 (mean value of +6.2 ). The total organic carbon (TOC) content of this sedimentary unit is high, between 3.5 and 8.8 wt.%, with δ 13 C TOC values (expressed in relative to VPDB) between −25.9 and −23.7  Table S2). Their TOC contents are highly variable with values between 0.6 and 17.9 wt.%, associated with δ 13 C TOC values between −33.3 and −21.7 (mean value of −27.8 ). The upper part of the FB 1c unit corresponds to a major upwelling event and Mn oxide precipitation during a deoxygenation episode (Fig. 2). Black shales of this sedimentary unit deposited in the upwelling zone have Se concentrations between 2 and 3 μg/g, but show higher, albeit near-to-zero, δ 82 Se values (between −0.2 and 0.0 ) relative to the underlying units ( Fig. 2; Table S2). Their δ 15 N values, exclusively above +5 (mean value of +6.2 , with no systematic offset between δ 15 N bulk and δ 15 N ker values), are also much higher than those for the underlying lower FB 1c and FB 1b units, but are in the same range as those for the transgressive FB 1a unit ( Fig. 2; Table S2). The upper FB 1c unit has TOC contents in the 4 to 6 wt.% range and δ 13 C TOC values between −32.5 and −31.6 (mean value of −32.1 ), which are slightly lower than those for the lower part of the FB 1 Member.
The FB 2 Member, deposited in a predominantly oxygenated seawater column, has been inferred to represent a return to an oxygenated surface ocean following the global deoxygenation event recorded by the underlying Mn-bearing Upper FB 1c unit (Ossa Ossa et al., 2018). Black shales of the FB 2a unit have Se concentrations between 2 and 3 μg/g, with near-to-zero δ 82 Se values at +0.2 , whereas δ 15 N shifts to near-zero values between −0.7 and +1.4 (mean value of +0.8 , with no systematic offset between δ 15 N bulk and δ 15 N ker values), relative to the significantly higher values (δ 15 N > +5 ) recorded in the underlying Upper FB 1c unit deposited during a deoxygenation event ( Fig. 2; Table S2). The TOC content is between 0.7 and 4.8 wt.%, with δ 13 C TOC values between −34.7 and −31.5 (mean value of −33.1 ). In the FB 2b unit, the δ 82 Se values are negative, between −0.5 and −0.4 in the black shales, with Se concentrations of 1-2 μg/g ( Fig. 2; Table S2). Early diagenetic pyrite concretions hosted in these FB 2b unit black shales yield higher Se concentrations, between 11 and 66 μg/g, with lighter δ 82 Se values between −1.1 and −0.2 (mean value of −0.8 ) ( Fig. 2; Table S2). The δ 15 N values are predominantly  (Ossa Ossa et al., 2013Préat et al., 2011). Fe HR /Fe T data for the FB Formation (empty circles) are from previous work (Canfield et al., 2013;Ossa Ossa et al., 2018), while data shown as black-filled circles for the Upper FC 1 Member are from this study. Oxic, anoxic and equivocal fields are determined according to Poulton and Canfield (2011) (see Methods in Supplementary Text SI 2). Dashed lines labeled with UCC on Mn/Al and Se/Al plots represent the values for upper continental crust (Rudnick and Gao, 2014). In the Se concentration and δ 82/76 Se plots, black-filled circles represent shale, grey-filled squares correspond to pyrite concretions, and empty, grey circles indicate previously published data from Kipp et al. (2017) and Mitchell et al. (2016). In the δ 15 N plot, black-filled circles are for bulk samples and grey-filled circles represent extracted kerogen. In the stratigraphic log numbers refer to the depositional setting (Dep. Set.; see Supplementary text): 1 = intertidal; 2 = upper shoreface to offshore transition; 3 = lower offshore; 4 = distal outer shelf. Shaded, gray fields represent the two-step deoxygenation event recorded by the Francevillian Group (Deoxy.) and associated with upwelling of anoxic deep-waters during deposition of the upper parts of the FB 1c unit and FC Formation (Ossa Ossa et al., 2018). Transgr. = transgression.
close to zero and range between −1.2 and +4.8 (mean value of +0.4 , with no systematic offset between δ 15 N bulk and δ 15 N ker values). The TOC and δ 13 C TOC values in the FB 2b unit are similar to those from the underlying FB 2a unit.
The last black shale unit analyzed in this study occurs in the Upper FC 1 Member, which corresponds to the end of the LE associated with a deoxygenation event (Ossa Ossa et al., 2018, 2021a.  Table S1).

Discussion
As for any ancient sedimentary rock evaluated for biogeochemical signatures and redox proxies, diagenetic and secondary alteration processes need to be carefully considered. The excellent preservation of the FB shales has already been indicated (Ossa Ossa et al., 2013) and further discussion can be found in the Supplementary Information (SI 1; Fig. S3). We further infer that postdepositional alteration processes had only a minimal effect on N isotope data presented in this study (see SI 3), as demonstrated by a lack of a systematic offset between δ 15 N bulk and δ 15 N ker values ( Fig. 2). Furthermore, the well-preserved pyrite concretions analyzed in this study rule out post-depositional alteration and/or surface oxidative weathering (Fig. S3). In addition, a significant effect on the Se isotope ratios of pyrite with such high Se concentrations seems unlikely, given the large amount of Se that would have had to be mobilized. Specifically for Se, the reducing conditions maintained by TOC and sulfide would have provided a strong O 2 -buffer, which would not allow much (if any) soluble, oxidized Se phases to be mobilized. In view of this, the excellent preservation, combined with reducing conditions, indicate that Se and N biogeochemical signatures are well preserved and thus robust.

Redox conditions in the Francevillian water column
The average Se concentration of the Francevillian black shales (1 to 7 μg/g) shows authigenic enrichment, with high Se/Al ratios between 0.04 and 1.43 ppm/wt.%, which are considerably higher than the average UCC value of 0.01 ppm/wt.% (cf. Rudnick and Gao, 2014). While Se appears to be mainly associated with sulfide phases in the Francevillian Group black shales (see Fig. S4), this enrichment is within the range of other Paleoproterozoic and late Neoarchean sedimentary successions, and is interpreted to result from oxidative terrestrial weathering of selenium-bearing miner-als, followed by Se sequestration into organic-and sulfide-bearing sediments (Kipp et al., 2017;Mitchell et al., 2016;Stüeken, 2017;Stüeken et al., 2015;von Strandmann et al., 2015).
The modern Se geochemical cycle is dominated by aerobic processes, with an average δ 82 Se value of approximately +0.3 in seawater ( Fig. S5A; Cutter and Bruland, 1984;Stüeken, 2017;Stüeken et al., 2015;von Strandmann et al., 2015). Terrestrial runoff represents the dominant Se input to the oceans (∼89% of Se), with an average δ 82 Se value within the igneous inventory range of between −0.3 to +0.3 , while ∼10% is contributed by volcanic activity and/or aerosols (Stüeken, 2017). Incorporation into organic particles (δ 82 Se ≈ +0.3 ) and sulfide minerals (δ 82 Se ≤ +0.3 ) in suboxic and anoxic/euxinic areas (∼45 and 7% of Se, respectively), as well as adsorption onto ferromanganese oxides (δ 82 Se ≈ +0.3 ) in oxic seawater (∼48% of Se), represent the main oceanic Se output channels (Stüeken, 2017). In the modern, fully oxygenated ocean, negative sedimentary δ 82 Se values reflect partial reduction of SeO x 2− (i.e., SeO 4 2− , SeO 3 2− and HSeO 3 − ) during diagenesis, although a similar signal may also be generated in oxygen-minimum zone (OMZ) settings overlain by oxygenated waters (Fig. 2B). Oxic water masses are enriched in dissolved Se and continuously supply Se oxyanions, which prevents quantitative Se reduction in anoxic pore-waters or in OMZ settings ( Fig. S5A; Kipp et al., 2017;Mitchell et al., 2016;Stüeken, 2017;Stüeken et al., 2015;von Strandmann et al., 2015). In a strongly redox-stratified water column, sediments record positive δ 82 Se values, similar to, or exceeding those, of local seawater, due to near-quantitative reduction where the rate of SeO x 2− supply lags behind its removal ( Fig. S5A; Kipp et al., 2017;Mitchell et al., 2016;Stüeken, 2017;Stüeken et al., 2015;von Strandmann et al., 2015 (Stüeken, 2017;von Strandmann et al., 2015). Such conditions may have been due to a higher rate of organic carbon sequestration causing extensive anoxia during early diagenesis, thus likely exhausting the availability of electron acceptors, and/or Se assimilation into biomass, perhaps due to an overall lower seawater SeO 4 2− reservoir during deposition. In the latter case, SeO 4 2− might have been more sensitive to mild redox changes under moderately suboxic conditions in the Francevillian basin seawaters than redox proxies based on Fe and S.
For the upper part of the FC Formation black shales, the δ 82 Se data show near-to-zero to slightly negative values, which may be interpreted in different ways (Stüeken, 2017). However, our new Fe speciation data indicate deposition under euxinic to anoxic, ferruginous conditions (Figs. 2, S2B; Table S1). Furthermore, previously published δ 82 Se data from a much higher resolution sample set from this sedimentary succession show predominantly positive to near-to-zero values and minor negative values ( Fig. 2; Kipp et al., 2017;Mitchell et al., 2016), which is consistent with an euxinic to anoxic/suboxic setting. Indeed, a strongly redox-stratified marine setting, with a shallow redoxcline, has been proposed for the upper part of the FC Formation ( (Zhang et al., 2014). Interpretations of N isotope data from ancient sedimentary rocks have been challenged on the basis that the N isotopic composition of the atmosphere may have changed over time (e.g., Jia and Kerrich, 2004;Kerrich et al., 2006). However, N isotope analyses of fluid inclusions in cherts have shown that the δ 15 N value of atmospheric N 2 has likely stayed close to 0 since at least ca. 3.5
The Francevillian Group black shales are dominated by δ 15 N values in the range of −2 to +3 (average +1 ± 3 ), with positive shifts of > 5 (having no relationship to water depth) at three stratigraphic levels ( Fig. 2; Table S2). The assumed low degree of post-depositional alteration of the N isotope data suggests that the δ 15 N values obtained for the studied FB and FC formations OM-rich shale and silty shale samples should provide a best estimate of the primary isotopic composition of middle Paleoproterozoic sediments and organic matter (see SI 3; Figs. 2, S8).
Combined Fe speciation, δ 82 Se, δ 34 S and trace metal concentration data indicate deposition under a well-oxygenated water column. By contrast, a few samples with elevated Fe HR /Fe T (> 0.38) ratios from the FB 2b unit, as well as the predominantly small δ 15 N values of +1 ± 3 in the FB Formation ( Fig. 2; Table S2), could be interpreted to reflect deposition under anoxic water-column conditions with a purely anaerobic N cycle. However, these elevated Fe HR /Fe T ratios and low δ 15 N values contrast with negative δ 82 Se values, low V/Al, low Mo/Al, and low U/Al ratios that are suggestive of oxic water column conditions (Ossa Ossa et al., 2021a;Fig. 2). We propose that the elevated Fe HR /Fe T ratios more likely represent upwelling of deep anoxic waters onto oxic shallow shelves (potentially changing the chemocline depth), causing precipitation of iron from the water column. Hence, considering this generally oxic setting, it is unlikely that these low δ 15 N values reflect a purely anaerobic N cycle. Instead, we interpret the N isotope data to reflect quantitative nitrification in an oxic seawater column, while denitrification mainly occurred in sediments, which typically results in very small isotopic effects (δ 15 N values of +1 ± 3 ).
This interpretation further suggests that the redoxcline (at which water-column denitrification occurs and imparts large fractionations in δ 15 N) was generally situated much deeper on the outer shelf (maximum water depth < 200 m; SI 1), or even below it, and that the Francevillian basin was not connected to a larger, isotopically fractionated marine NO 3 − reservoir during deposition of the FB and FC formations. One possible explanation might be that the basin was restricted at this time. However, this seems unlikely because the FB and FC formations record the LE, which was a typical feature of the open ocean (Bekker, 2022;Karhu and Holland, 1996) pool was evidently large enough to be assimilated into biomass followed by subsequent preservation in the sediments (Figs. 2, 3). This is further supported by previously published N isotope data for the upper part of the FC Formation (δ 15 N > +5 ), indicating partial water-column denitrification and/or anammox, with a NO 3 − pool that was large enough to leave an isotopic signature of strong redox cycling of nitrogen in the water column (Kipp et al., 2018). Importantly, deposition of the upper parts of the FB 1c unit and FC Formation coincided with a global two-step deoxygenation event during a period of enhanced submarine volcanic activity at the end of the LE (Ossa Ossa et al., 2018). Associated with these phenomena, an episodic redoxcline encroachment onto the inner shelf would have locally provided anoxic conditions for partial denitrification and/or anammox in the water-column, which is supported by highly negative δ 13 C TOC values (down to −48 , V-PDB), reflecting enhanced activity of heterotrophs (e.g., methanotrophs and denitrifiers) in the water column during these two deoxygenation events (Figs. 2, 3; Table S2; Ossa Ossa et al., 2018). Before these events, the NO 3 − reservoir progressively grew as a result of enhanced oxygenic photosynthesis during the LE, which resulted in an expansion of oxygenated conditions in the Francevillian basin from the shelf edge to outer shelf. This scenario implies availability of free O 2 , at least at the level of a few μM, to allow buildup of NO 3 − above the redoxcline. Nitrate reduction becomes pervasive when dissolved oxygen is below ∼2 μM (Lam and Kuypers, 2011), which is equivalent to roughly 1% of modern seawater dissolved oxygen in equilibrium with the atmosphere. Hence, seawater dissolved oxygen at the depositional site of the upper parts of the FB 1C unit and FC Formation appears to have been above this threshold. Our combined N and Se isotope data thus provide a minimum level of 1% modern marine O 2 saturation level for the dissolved oxygen content of seawater at the end of the LE. Another sedimentary succession, the lower ca. 2.1-2.0 Ga Zaonega Formation of Russia (see Ossa Ossa et al., 2018Ossa et al., , 2021a for detailed discussion on the depositional age), where Fe speciation and trace-metal data suggest deposition under oxic conditions during the late stage of the LE (Canfield et al., 2013;Scott et al., 2014), also shows a correlation between extracted kerogen δ 15 N values < +2 (Kump et al., 2011) and light δ 82 Se values ( Fig. 4; Table S3).
By contrast, the upper part of the Zaonega Formation, which was deposited during the deoxygenation event at the end of the LE (Asael et al., 2018;Canfield et al., 2013;Kump et al., 2011;Scott et al., 2014), yielded more positive δ 15 N values of > +2 for extracted kerogen (Kump et al., 2011) and more positive δ 82 Se values ( Fig. 4; Table S3). Combined, the Francevillian Group and the lower Zaonega Formation N and Se isotope data show parallel trends associated with similar seawater redox changes. Continental-shelf black shales with near-to-zero δ 15 N values of extracted kerogen are also known from several other sedimentary successions recording the LE, including the ca. 2.2 Ga Wewe Slate, ca. 2.15 Ga Sengoma Argillite Formation, and ca. 2.1 Ga Hautes Chute Formation (Kipp et al., 2018). Collectively, this indicates that oxic conditions with small seawater NO 3 − -and SeO 4 2− -reservoirs, episodically affected by an encroaching redoxcline, were a common feature in marginal-marine basins across the LE until its end. Hence, enhanced biological O 2 production drove pervasive oxygenation in shelf environments, which likely shifted the redoxcline deeper in the water column to allow growth of the O 2 , NO 3 − and SeO 4 2− pools in open, marginal-marine settings during the LE (Fig. 3).

Implications for Earth surface oxygenation during the LE
Based on the molybdenum isotope records of black shales deposited during the GOE, it has been inferred that the oceans were redox-stratified and largely anoxic, but with pulsed intervals of progressively expanding euxinia from the early stage of the GOE (ca. 2.43 Ga) to its termination (ca. 2.06 Ga) (Asael et al., 2018). This view is also supported by previously published Se isotope data for contemporaneous continental-shelf black shales, which are dominated by positive δ 82 Se values (Fig. 5;Kipp et al., 2017;Mitchell et al., 2016;Stüeken et al., 2015), inferred to reflect quantitative SeO x 2− reduction in strongly redox-stratified oceans with a shallow redoxcline situated above storm wave-base (Fig. 5;Kipp et al., 2017). The predominantly positive δ 15 N values (> +4 ) in these black shales deposited during the early stage of the GOEbefore the LE-are also consistent with partial water-column denitrification and anammox in strongly redox-stratified oceans with a significant nitrate reservoir in the oxic surface layer (Cheng et al., 2019;Kipp et al., 2018;Stüeken et al., 2016), similar to that inferred for the two-step deoxygenation event at the end of the LE recorded in the Francevillian basin (Figs. 2, 3; Table S2). It thus appears that positive δ 15 N values recorded in marginal-marine settings at the beginning and end of the GOE-before and immediately after the LE-reflect a locally enhanced isotopic expression of partial water-column denitrification and/or anammox, as a consequence of redox stratification. A minimum O 2 surface ocean concentration of >0.4 μM has been proposed during these two time intervals, based on Se isotope data alone (Kipp et al., 2017). Our combined N and Se isotope data increase the minimal seawater dissolved O 2 concentration to >2 μM for the later timer interval.
Molybdenum isotope data show that during the LE, the ocean redox state evolved towards the more expansive development of oxic conditions, before returning at the end of the LE to oceans with a shallow redoxcline and extensive development of euxinic conditions (Asael et al., 2018). High I/(Ca+Mg) ratios in shallowwater carbonates, which are interpreted to reflect a substantial IO 3 − reservoir in the surface ocean, with dissolved O 2 concentrations of at least 1 μM, agree with more expansive seawater oxygenation during the LE (Hardisty et al., 2014), although offshore environments-outer shelf and beyond-are inferred to have remained anoxic (Kipp et al., 2017;Stüeken et al., 2016) gested for proximal-shelf marine environments during the early stage of the GOE (Cheng et al., 2019;Kipp et al., 2018;Stüeken et al., 2016), and such conditions might have developed, at least episodically, as early as the Neoarchean (cf. Stüeken et al., 2016 and references therein). However, this NO 3 − pool did not reach a pervasive and stable level in offshore environments situated on the lower part of the outer shelf during the Paleoproterozoic (Cheng et al., 2019;Kipp et al., 2018;Stüeken et al., 2016). Our combined N and Se isotope data show for the first time that the NO 3 − and SeO 4 2− reservoirs were persistent and stable across the entire continental shelf during the LE (Fig. 3). This requires a much higher rate of oxidant accumulation, with O 2 and NO 3 − concentrations well above the minimum levels reached during the early stage and end of the GOE, which indicates enhanced biological O 2 production in the surface ocean. Such a high rate of oxygenic photosynthesis in shallow-marine environments must have caused O 2 downwelling, resulting in a deeper position of the redoxcline in the water column (Fig. 5), which nevertheless likely remained shallower than in the NOE oceans (Alcott et al., 2019;Canfield et al., 2007;Lenton et al., 2014;von Strandmann et al., 2015). Thick sulfate evaporite deposits, consistent with high seawater sulfate concentrations and developed in a supposedly marginalmarine basin during the LE, have been linked to the buildup of an oxidant reservoir equivalent to more than 20% of the modern atmosphere-ocean oxidizing capacity (Blättler et al., 2018). However, an open-marine setting for this succession in the Onega basin on the Karelia craton has been recently questioned on sedimentological and geochemical grounds (Alfimova et al., 2022), and an apparently large thickness of the sulfate evaporite deposit might reflect a drill core intersection through the central part of a salt diapir rather than true stratigraphic thickness (cf. Esipko et al., 2014). Furthermore, based on large U and Cr isotope fractionations recorded in black shales deposited at the end of the LE, it has been proposed that the middle Paleoproterozoic oceans remained well-oxygenated even after the end of the LE, and that modern-style biogeochemical cycling developed during the middle Paleoproterozoic (Mänd et al., 2020(Mänd et al., , 2022. However, the Cr and U isotope data can be explained by slow depositional rates and the removal of redox-sensitive elements from the oceans under sub-oxic to weakly euxinic settings after the GOE. Furthermore, large Cr and U isotope fractionations, which have not been systematically recognized in sedimentary deposits until the Neoproterozoic (e.g., Dang et al., 2022;Planavsky et al., 2014), may alternatively suggest that these redox-sensitive elements were drawn from a relatively small seawater reservoir, either in the whole ocean or in somewhat isolated basins (see Andersen et al., 2020;Dang et al., 2022). Considering the regional stratigraphic framework for Karelia in Finland and Russia, where multiple sedimentary basins record flooding at the end of the Lomagundi carbon isotope excursion, as also recorded by the Zaonega Formation, it appears that these large U and Cr isotope variations over short stratigraphic intervals likely reflect a very small U and Cr reservoir in the ocean. It is also important to emphasize that multistage reduction-oxidation processes subsequent to deposition and in association with hydrothermal fluid or hydrocarbon migration, as well as a wide range of redox-independent processes, could fractionate Cr and U isotopes regardless of ocean dissolved oxygen content (Andersen et al., 2020;Babechuk et al., 2017Babechuk et al., , 2018Dang et al., 2022;Miletto et al., 2021), which limits their application to infer O 2 levels in the atmosphere-ocean system.
Regardless of this difference in interpretation, in contrast to N and Se, S and U yield oxyanions that are stable at low Eh (< 0 V; Figs. S6, S7; Rue et al., 1997;Stumm and Morgan, 1970) under the circumneutral pH conditions of the Paleoproterozoic oceans (Halevy and Bachan, 2017). Although our combined N and Se isotope data support extensive ocean oxygenation during the LE, it appears that the NO 3 − inventory was not sufficient to shift the N isotopic composition of the global ocean. A model arguing for biogeochemical cycling comparable with the oxidant reservoir approaching the size of the modern atmosphere-ocean capacity (Blättler et al., 2018;Mänd et al., 2020) thus seems unlikely for the middle Paleoproterozoic. Moreover, the N and Se isotope data indicate that the oxyanions of these two elements, which are redoxsensitive at a higher redox potential (Eh > 0.4 V at circumneutral pH; Figs. S6, S7, S1), were more attuned to O 2 dynamics in the atmosphere-ocean system across the GOE, compared to S-and Ubased proxies. Indeed, N and Se isotope data record progressive oxygenation of the atmosphere-ocean system from the beginning  (Bekker, 2022;Bekker et al., 2003Bekker et al., , 2016Bekker and Holland, 2012;Karhu and Holland, 1996;Ossa Ossa et al., 2018;Planavsky et al., 2012). The lower panel shows the distribution of oxic (blue) vs. anoxic (grey) conditions in the water column.
(1) = predominantly anoxic ocean with oxygen oases in sublittoral zone (e.g., Lyons et al., 2014;Kipp et al., 2017;Stüeken et al., 2016); (2, 4) = redox-stratified oceans with the maximum depth of the redoxcline above the storm wave-base; (3) = redox-stratified ocean with a relatively deep redoxcline in the distal outer shelf or below. Empty circles are δ 82/76 Se data from previous studies (Kipp et al., 2017;Mitchell et al., 2016;Stüeken, 2017;Stüeken et al., 2015) (δ 82/76 Se = 1.5*δ 82/78 Se). Filled circles represent δ 82/76 Se data from this study. Shaded, grey field represents δ 82/76 Se range of modern seawater. (For interpretation of the colors in the figure(s), the reader is referred to the web version of this article.) of the GOE to the LE, which is not clearly expressed in S and U isotope data (e.g., Blättler et al., 2018;Holland, 2006;Mänd et al., 2020;Ossa Ossa et al., 2018;Planavsky et al., 2012;Scott et al., 2014). Furthermore, N and Se isotope data also record widespread ocean deoxygenation at the end of the LE/GOE that is not obvious in U and Mo isotope records (Mänd et al., 2020; but see Asael et al., 2018 andAndersen et al., 2020 for a different view on secular Mo and U isotope variations). Development of deoxygenated marine conditions at the end of the LE is further supported by high V enrichment in black shales (Asael et al., 2018;Ossa Ossa et al., 2018;Scott et al., 2014), which reflects strong anoxia in open, continental-shelf settings (Canfield et al., 2013;Ossa Ossa et al., 2021a). Anoxic seawater conditions at the end of the LE does not imply that these oceanic conditions remained steady until the NOE, and indeed, several local to potentially global mid-Proterozoic oxygenation events have recently been described (Canfield et al., 2018;Dang et al., 2022;Luo et al., 2021;Stüeken et al., 2021;Zhang et al., 2018). The evolving picture of oxygen dynamics in the Paleoproterozoic atmosphere-ocean system thus highlights an increase in the amplitude of oxygenation from the beginning of the GOE to the LE, with superimposed smaller scale oxygen oscillations. This implies that, despite transiently elevated oxygen concentrations during the LE, feedbacks linking terrestrial weathering, nutrient controls on primary organic productivity, and low oxygen levels maintained the atmosphere-ocean system at an oxygenation state far below the modern level.

Conclusions
Dissolved nitrate was heterogeneous and unstable in offshore marine environments during the ca. 2.43-2.06 Ga GOE. By contrast, new combined Se and N isotope data presented here show that continental shelf waters were highly oxygenated, with unlimited supply of nitrate and selenium oxyanions during the peak of the LE. Although enhanced accumulation of oxidants clearly demonstrates high rates of oxygenic photosynthesis in shallowmarine environments during the LE, the oceanic O 2 inventory remained well below the modern level. Pervasive redox stratification in the deep oceans during the LE implies that the global extent of modern biogeochemical cycles and their isotopic effects were not achieved until the end of the Neoproterozoic.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Africa. We acknowledge the helpful comments by Kaarel Mänd, an anonymous reviewer and the handling editor, Boswell Wing, which greatly helped to increase the quality of the manuscript. We also thank Johnathan Mboulou Ella for his tremendous logistic support in the field.