Earth's atmosphere is suggested to have shifted from anoxic to oxic in early Palaeoproterozoic during ~2.45–2.2 billion years ago (Ga)1,2,3,4,5,6,7,8. Any increase in O2 would have resulted in reduced atmospheric methane levels, triggering glaciations2. Alternatively, dramatic geological events such as a global glaciation could have caused a shift in the balance of oxidants and reductants at Earth's surface8,9,10. Although several hypotheses have been proposed with regard to correlating geologic events and atmospheric transition, their relationship remains poorly constrained. To develop a more complete history of interactions in Earth's system, investigating continuous records of atmospheric O2 levels in response to geological events is prerequisite.

The abundance of redox-sensitive elements in sedimentary rocks is a useful proxy of atmospheric O2 levels5,6,11. Under an oxidizing atmosphere, the redox-sensitive elements in continental crust (for example, Os, Re and Mo) are mobilized in the hydrological cycle by oxidative weathering, leading to high concentrations in sediments deposited within anoxic oceans11. Under an anoxic atmosphere, in contrast, redox-sensitive elements do not enter the oceans through continental weathering11. We propose that records of the abundance of the redox-sensitive element Os and its initial 187Os/188Os values in sedimentary rocks represent a promising tool for reconstructing variations in atmospheric O2 during the Great Oxidation Event. Continental sulphides and granites contain high levels of radiogenic 187Os (currently, 187Os/188Os~1.4)12,13. Accordingly, onset of the oxidative delivery of continental Os to the oceans would be recorded as a rise in 187Os/188Os value in sediments12 relative to the value dominated by non-radiogenic mantle-derived and extraterrestrial sources of Os (187Os/188Os≈0.11 in the Palaeoproterozoic)12. In addition, because the redox potential for the oxidization of Os is higher than that of conventional redox-sensitive elements (Mo and Re), the records of Os abundance in sedimentary rock may allow us to reconstruct the accumulation of high levels of atmospheric O2. In fact, studies of the Os isotopic compositions of 2.50-Ga organic-rich shales11 and of authigenic pyrite in 2.32 Ga shales14 show no clear evidence of the oxidative weathering of Os, but Mo and Re enrichments are found in the 2.50-Ga organic-rich shale11.

In the present study, we analysed the abundances of redox-sensitive elements (Os, Re, Mo and U) and Os isotopic compositions in sediments deposited during and after the second Palaeoproterozoic glaciation, to evaluate the role of climate change in the rise of atmospheric O2. We focus on shallow-marine sediments, because the signal of the onset of oxidative weathering of Os would first appear in shallow-marine environments15,16. We found the enhancements of Re, Mo and Os with high initial 187Os/188Os in the sediments, suggesting that the oxidative delivery of these elements to shallow-marine environments has started in the period of climatic recovery from the second Palaeoproterozoic glaciation.


Sampling and the geological setting

The Huronian Supergroup is a sequence of early Palaeoproterozoic sediments and minor volcanic rocks. The studied samples were collected from the Bruce and Espanola formations of the 2.45–2.22 Ga Huronian Supergroup1,17 (Fig. 1). Presences of detrital pyrite and uraninite in the lower units of the Huronian Supergroup suggest low O2 levels in the atmosphere at the time of deposition1; meanwhile, both presence of redbeds and disappearance of mass-independent fractionation in sulphur in the upper units suggest that a significant rise in atmospheric O2 would have occurred around the second glaciation in the Huronian Supergroup1,8 (Fig. 1a).

Figure 1: Stratigraphy of the Huronian Supergroup.
figure 1

(a) Simplified stratigraphy of the Huronian Supergroup1, showing proxy indicators of redox conditions in the atmosphere1,8. The depositional age of the Huronian Supergroup is constrained to between about 2.45 and 2.217 Ga, based on the dating of dyke swarms in basement rocks and the post-Huronian Nipissing diabase, respectively1,2. Detrital uraninite and pyrite and mass-independent fractionation in sulphur (MIF-S) are found in the lower units1,8; whereas, presence of redbeds and disappearance of MIF-S are observed in the upper units1,8. (b) Stratigraphy of the boundary between the Bruce and Espanola formations. The Bruce Formation contains granitic and mafic dropstones of up to cobble size, in a matrix of grey-green sandstone. The boundary is defined as the uppermost occurrence of dropstones.

The sampling site of the present study is located in the western part of the Huronian Supergroup (called the Elliot Lake area) (Supplementary Fig. S1), where the boundary between the Bruce and Espanola formations is exposed on the east side of Highway 108 (refs 17,18) (46.248°N, 82.402°W). At the sampling site, a sandstone–siltstone interval (~1.5 m thick) separates diamictites of the underlying Bruce Formation from laminated limestone of the overlying Espanola Formation18 (Fig. 1b). The relatively sudden change from glacial diamictite of the Bruce Formation to carbonate of the Espanola Formation implies a climatic shift from icehouse to greenhouse conditions. The lower part of the sandstone–siltstone interval consists of massive sandstone that changes upward to laminated siltstone, probably reflecting upward deepening associated with a post-glacial marine transgression in a shallow-marine environment18. Individual and clustered authigenic sulphide crystals with diameters of ~100 μm or less are observed parallel to laminae in the sandstone–siltstone interval (Supplementary Fig. S2; Supplementary Note 1).

We analysed isotopic compositions of Os and Re, abundances of Os and other redox–sensitive transition elements, Re, Mo and U in the bulk rock samples (Methods). Total sulphur (TS) and total organic carbon (TOC) contents and carbon isotopic ratio in TOC (δ13Corg) in the bulk rock samples are also analysed (Methods). The surfaces of the whole rock samples were cut off and polished with corundum powders to remove potential contaminants. Then, the rocks were cleaned with ultrasonic. The cleaned samples (~2–4 g) were crushed with an agate mortar. For diamictite samples, we removed dropstones and used the sandstone-matrix. The regional settings and rock description are also described in Supplementary Notes 1 and 2. The results of analyses are summarized in Supplementary Tables S1-S3.

Geochemical trend of the Bruce-Espanola boundary

Figure 2 shows variations in the abundances of Os (non-radiogenic 192Os), Re, Mo and U across the boundary between the Bruce and Espanola formations (herein referred to as the Bruce–Espanola boundary). In the sandstone–siltstone interval, 192Os abundance is enriched by a factor of 2–20 relative to levels in typical upper continental crust and oceanic basalt. We used chromium-sulphuric acid and inverse aqua regia as dissolution media in Re–Os analyses (Methods and Supplementary Note 3). Given that chromium-sulphuric acid dissolves detrital components much less efficiently than does inverse aqua regia, the similarity of the results obtained using the different media (Fig. 2) indicates that the enrichment in Os (and Re) is not related to the detrital component. In addition, the Os concentration in cobbles (which is representative of the average Os concentration of the detrital component) in the diamictites is very low (192Os≈4 p.p.t.) compared with that in the bulk sediments (192Os~10–300 p.p.t.). These results indicate the authigenic enrichment of Os in the sediments. The sandstone–siltstone interval is also marked by high levels of Re and Mo (Fig. 2), which is consistent with the oxidative weathering of redox-sensitive elements and subsequent incorporation in the sediments19. Given the correlation between Os abundance and TS content in the sediments (Fig. 2; Supplementary Fig. S3), a large component of the Os in the sandstone–siltstone interval is hosted by authigenic sulphide minerals (Supplementary Note 2). These suggest that such minerals are good indicators of the Os isotopic composition of post-glacial bottom water and pore water in sediments.

Figure 2: Geochemical trends across the Bruce–Espanola boundary.
figure 2

Initial 187Os/188Os values, abundances of redox-sensitive elements, TOC content, carbon isotopic values in TOC (δ13Corg), and total sulphur content are shown. Light grey area represents the sandstone–siltstone interval. The average 192Os concentrations in upper continental crust12 and oceanic basalt20 are shown as dark and light lines, respectively, in the plot showing Os concentration. Dashed lines show average Re, Mo and U concentrations normalized by Al2O3 in upper continental crust34. Initial 187Os/188Os values were obtained based on a depositional age of 2.3±0.2 Ga. The grey line in the plot showing initial 187Os/188Os represents the value of mantle-derived components at the time of deposition20. The errors in the initial 187Os/188Os values are mainly derived from the errors in depositional age. Black circles in the 192Os and Re plots represent the results using inverse aqua regia dissolution media, and red and blue circles represent those using chromium-sulphuric acid.

Based on a depositional age of 2.3±0.2 Ga (an age bound of the Huronian Supergroup), we calculated temporal variations in initial 187Os/188Os around the Bruce–Espanola boundary (Fig. 2). The initial 187Os/188Os shows an abrupt increase to ~0.8 at the base of the sandstone–siltstone interval, and varies between 0.4 and 1.1 within the interval; the value within the diamictite is very low, similar to the value of mantle-derived components at ~2.3 Ga (ref. 20). The increase in initial 187Os/188Os in the sandstone–siltstone interval is matched by an increase in Os abundance, strongly indicative of a marked increase in the input of radiogenic Os at the cessation of glaciation. Elevated Os levels in sedimentary rocks might also arise from extraterrestrial21,22 or hydrothermal inputs23; however, these processes typically provide non-radiogenic Os21,23. The occurrence of much younger alteration would have added Re and radiogenic Os via interaction between oxygenated fluids and organic material in the sediments. However, such interaction would have resulted in a large degree of scatter in the Re–Os isochron plot. Although the scatter in the isochron plot may represent small-scale (<~10 cm; that is, the sample resolution) mobilization of Os and Re by post-depositional alteration, all the data plot within the range expected to develop over a period of 2.3 Ga given initial 187Os/188Os values that are intermediate between mantle-derived and continental values (Fig. 3; Supplementary Note 4). These results do not support the much later addition of radiogenic Os.

Figure 3: Re–Os isochron diagram for the samples.
figure 3

The grey area represents the range of 187Re/188Os and 187Os/188Os expected to develop over ~2.3 Ga given initial 187Os/188Os values of between the mantle-derived (0.11) and continental values (~1.4)12. Colours of the data points represent the positions in the stratigraphy. Red circles represent the samples from the Bruce formation (<0 m from the boundary). Blue squares and green diamonds represent the samples from the lower (0–0.5 m from the boundary) and upper part (0.5–1.5 m from the boudanry) of the sandstone–siltstone interval, respectively.


Given the above observations, we interpret the systematic variations in Os record observed around the Bruce–Espanola boundary to represent evidence for the onset of the oxidative weathering of radiogenic Os on the continents12 (Fig. 4). We hypothesize that atmospheric O2 levels increased immediately after the second Palaeoproterozoic glaciation sufficient to mobilize continental Os into the hydrological cycle. Because a large part of the oceans was possibly still anoxic at that time5,24, the dissolved Os delivered via rivers would have been reduced to immobile Os over a short period, probably before mixing homogeneously in the oceans. In such case, most of the delivered Os would have accumulated within shallow-marine environments (Fig. 4; Supplementary Note 5). Within deep-marine environments during the same period, such as the lower Timeball Hill Formation25, in contrast, continental Os would have remained depleted because of the short residence time of Os in the anoxic ocean. This would have resulted in the non-radiogenic, hydrothermal Os in the Timeball Hill Formation14.

Figure 4: Redox conditions of the atmosphere and oceans.
figure 4

(a) During the second Palaeoproterozoic glaciation, the atmospheric O2 levels are likely to have been low. It is considered that the shallow-marine environments were also reducing during the glaciation. (b) In the glacial aftermath, the atmospheric O2 levels would have become high, sufficient to deliver radiogenic Os to the ocean. The redox-sensitive elements on continents, such as Os, Re and Mo, would have been delivered to the ocean under oxidizing atmospheric conditions and then would have been reduced and deposited in shallow-marine environments, leading to enrichments in the sediments.

We estimate the O2 level required for continental Os to enter the oceans, above the level at which Fe2+ is oxidized to Fe3+ in soils, as Fe2+ has a major role as a reductant in converting dissolved Os into immobile Os26. Thus, if Fe2+ is abundant in rivers, little Os would be expected to enter the oceans. Under anoxic atmospheres, soils are a source of Fe2+ to rivers. To oxidize Fe2+ to Fe3+ in soils, atmospheric O2 should be ≥~10−4 times the present atmospheric level (PAL)5,6; accordingly, O2 is considered likely to have reached this level. Insoluble U4+ can be converted to soluble U6+ under conditions similar that for oxidization of Fe2+ to Fe3+ (ref. 27) Although a gradual, upward increase in U abundance (Fig. 2) might reflect an increase in soluble U6+ concentration, low levels of U enrichment relative to those of Re, Mo, and Os imply that atmospheric O2 would not have substantially exceeded ~10−4 PAL. Figure 2 also shows that the observed high 187Os/188Os values seem to decrease in the upper part of this section, whereas those of Mo and Re do not. This also implies that the atmospheric O2 had been fluctuated around ~10−4 PAL where Mo and Re is sufficiently delivered to the oceans but Os is not.

Previous studies have proposed a possible link between glaciations and the rise in O2 (refs 2,3,8,9,10). For example, a global warming in a glacial aftermath may have provided nutrients to the ocean via chemical weathering, leading to a cyanobacterial bloom8,9,17. In fact, the recovery of carbon isotopic values in TOC to ~25‰ preceding Os enrichment (Fig. 2) implies that primary production by photosynthesis has become predominant in the glacial aftermath (Supplementary Note 6). In addition, melting at the glacier base would release H2O2, which interacts with seawater to form O2 (ref. 9). Our observations support the hypothesis that climatic recovery acted to accelerate atmospheric oxidization3,8,9,10.

Alternatively, if the rise in O2 occurred before the second Palaeoproterozoic glaciation, because of increased subaerial volcanism28, another explanation is required for the low abundances of redox-sensitive elements and low initial 187Os/188Os in the Bruce Formation. The Snowball Earth hypothesis3,29 predicts a shutdown of the hydrological cycle, thereby severely restricting the input of redox-sensitive elements into the oceans. To assess the possibility of a Snowball Earth at the time of the glaciation recorded by the Bruce Formation, further palaeomagnetic studies are required. If that is confirmed, it would mean that the Bruce Snowball glaciation preceded the Makganyene Snowball glaciation2,3 in the Palaeoproterozoic, and that Snowball Earth events might have occurred as pairs at both the beginning and end of Proterozoic in Earth's history.


Re-Os analysis

Sample weights for Re-Os analysis are ~1.0–2.0 g for diamictite and ~0.3 g for the sandstone–siltstone interval. The analytical procedure for the Os and Re concentrations and Os isotopes of this study is based on isotope dilution combined with Carius tube digestion, carbon tetrachloride and solvent extractions, microdistillation, and column separations (for example, refs 20,30). For each analysis, powdered whole rock samples, mixed with a known amount of 185Re and 190Os spike solutions, were dissolved and equilibrated with dissolution medium using the Carius tube method. We used two types of dissolution medium for the Re-Os analysis; chromium-sulphuric acid (5 ml) and inverse aqua regia (10 ml). The chromium-sulphuric acid had to be purified for Re and Os before use. The purification method given by ref. 31 was followed in the present study. After heating sealed Carius tubes at 230 °C for at least 24 h, Os was extracted by carbon tetrachloride and subsequently back-extracted into 9 N HBr. Micro-distillation further purified the extracted Os. Rhenium was separated from the aqueous phase using Muromac's AG 1-X8 anion exchange resin30,32.

The Os and Re isotopic measurements were performed with a ThermoFinnigan TRITON (NTIMS) at JAMSTEC30,32. The Os was loaded on a high-purity Pt filament that had been baked in air to high temperatures for 3 min before loading. The samples were covered with 10 μg Ba using a 10,000 p.p.m. Ba(NO3)2 solution. The instrumental mass fractionation of Os was corrected by normalizing the measured 192Os/188Os ratio to 3.08271 (ref. 30). Rhenium isotope ratios were measured with a total evaporation technique32. The total blank levels for Re and Os are shown in the note of Supplementary Tables S1 and S2. The reproducibility of replicate analyses was examined using molybdenite (high Os concentrations) and andesite (low Os concentrations) by ref. 30. The Re-Os model ages determined for the standard molybdenite samples have an excellent reproducibility of ±0.34% (n=3, 2 σ)30. The Re-Os age for the standard molybdenite samples (221.1±0.8 Ma) is also in excellent agreement with the Re-Os ages of 221.5±0.3 and 221.3±0.2 Ma obtained previously30. The 187Os/188Os ratios and Re and Os concentrations of the five standard andesite samples are 0.2092±0.0078 (2 σ) and 46.7±1.6 and 11.4±1.5 p.p.t., respectively30. The reproducibility of the standard andesite samples is also high enough for Re-Os analyses of low-Os samples. More detailed data about the reproducibility are shown in ref. 30.

Major and trace elements

Bulk chemical and mineralogical compositions of the rock samples were analysed with standard X-ray fluorescence technologies (AXIOS, PANalytical) and an X-ray Diffractometer (X'PERT-PRO, PANalytical), respectively, at University of Tokyo. For the measurements of Mo and U abundance, we used ~100 mg of the rock powder. The powder samples were digested with HNO3, HClO4 and HF. After evaporation of acids to near dryness, the residues were dissolved in a HNO3 (2%) and HF (0.1%) solution. The Mo and U abundances in the solutions were measured with an inductive-coupled plasma mass spectrometer (7500ce; Agilent) at JAMSTEC.

Sulphur analysis

TS analysis for the powder samples was conducted with a helium-gas continuous-flow isotope-ratio mass spectrometer (Isoprime-EA (Isoprime)) at University of Tsukuba. Sample weights for the TS analyses are ~3 mg. TS contents of the samples represent the mean values of four repeated measurements of the powder samples. Errors for the TS contents are obtained from one standard deviation of the repeated measurements.

Carbon analysis

For TOC and δ13Corg analyses, the powder samples were treated with 4 M hydrochloric acid for 2 days. Then, they were washed with distilled water several times to remove calcium salt and dried in an oven. Measurements for TOC contents and their carbon isotopic compositions were performed by an on-line system of Finnigan Delta Plus XP isotope-ratio mass spectrometry at JAMSTEC33,35. Total organic carbon content in the samples was determined by calibrated ion currents of m/z 44, 45, and 46 in isotope ratio mass spectrometry. Carbon isotopic compositions were expressed as conventional δ notation against V-PDB standard; δ13C=(13Rsample/13Rstandard−1)×1,000, where 13R is 13C/12C ratio.

Additional information

How to cite this article: Sekine, Y. et al. Osmium evidence for synchronicity between a rise in atmospheric oxygen and Palaeoproterozoic deglaciation. Nat. Commun. 2:502 doi: 10.1038/ncomms1507 (2011).