Nitrogen isotopic techniques for the Namibian samples are presented in (Johnson et al., 2017b), and are similar for Mineral Fork and Kelley Canyon Formation samples. Nitrogen concentration and isotope ratios for the Mineral Fork Formation were characterized at the University of Colorado Boulder Earth Systems Stable Isotope Lab (CUBES-SIL), and values for the Kelley Canyon Formation were determined, using identical protocol, at the Earth System Evolution Lab (EaSEL) at Iowa State University. Carbon concentrations were obtained during the same runs as N concentration and isotopic analyses, using the same standards.

Sample powders were decarbonated via reaction with 5 ml of 6 N HCl. Samples were then sonicated for 30 min. All tubes were then placed in a 60°C oven overnight. The next day, samples were centrifuged to settle all undissolved material. Acid was poured off, fresh acid was added as before, and samples sat in the oven overnight. This acid refresh was repeated once more. To clean samples, all were rinsed three times with DI H₂O, centrifuging between each rinse. Sample powders then dried at 60 °C for 2 days; all vials containing multiples of the same sample powder were combined and homogenized after drying.

Samples were then analyzed on a Thermo Delta V after combustion in a Thermo Elemental Analyzer at CUBES-SIL and on a Thermo Delta V Plus following combustion in a Thermo Isolink EA at EaSEL. Between 50 and 100 mg of sample powder was weighed into a Sn capsule, as well as standards: two acetanilides (act1, δ ¹⁵N = 1.18⁰/₀₀ ± 0.02 and act2, δ ¹⁵N = 19.56⁰/₀₀ ± 0.03) and pugel (δ ¹⁵N = 5.45⁰/₀₀ ± 0.1) at CUBES-SIL and urea (internal, δ ¹⁵N = −2.60 ± 0.2⁰/₀₀) and caffeine (USGS62, δ ¹⁵N = 20.17 ± 0.1⁰/₀₀) at EaSEL. All samples were flash-combusted with an excess of O₂ at 1,020° C in a combustion column packed with cobaltous oxide (combustion aid) and silvered cobaltous oxide (sulphur scrubber). Combustion products were passed over a reduced copper to reduce all N to N₂ and absorb excess O₂. Finally, sample gas was passed through a magnesium perchlorate trap to absorb water and a 3 m gas chromatography column to separate N₂ from CO₂. All analyses were quantified using IsoDat software and the Isoreader pipeline (Kopf et al., 2021). Errors reported are standard deviations from repeated analyses.

While the dominant mineralogy of the Ghaub at this location is carbonate, the geochemical results presented here, and in Johnson et al. (2017b), reflect analysis of the siliciclastic portion of the Ghaub sediment. Carbonate was removed via dissolution, so that the clay minerals could be analyzed directly. Clays record both the N signal (Ader et al., 2016) and the trace metal budget (Tribovillard et al., 2006; Johnson et al., 2017b).

Mineral Fork and Kelley Canyon samples were analyzed in the same run as previously measured glacial till samples (Timeball Hill and Blaubeker) from published work (Gaschnig et al., 2016; Johnson et al., 2017b). Measured N and δ ¹⁵N for Timeball (250 ppm, 5.06⁰/₀₀) and Blaubeker (158 ppm, 4.1⁰/₀₀) are similar to published values: 305 ± 5 ppm and 4.9 ± 0.3⁰/₀₀ for Timeball Hill; and 90 ± 20 ppm and 4.4 ± 0.2⁰/₀₀ for Blaubeker. We note that agreement in δ ¹⁵N values is quite good, though there are differences in N concentration outside uncertainty. Further, we also ensured that organic standards (urea, acetanilide, pugel, and caffeine) used to calculate N concentration included very small standard amounts ( < 10 μ g total mass) to capture relatively low N content in these rocks. The overall N concentrations are in the 100s of ppm, which is a reasonable N content suitable for EA-IRMS analysis (Bräuer and Hahne, 2005; Boocock et al., 2020).

The Fe speciation method targets operationally defined Fe pools, including carbonate-associated Fe (Fe_carb), ferric oxide Fe (Fe_ox), magnetite Fe (Fe_mag) and pyrite Fe (Fe_py). Extractions were performed according to well-established protocols (Poulton and Canfield, 2005, 2011; Poulton, 2021), with subsequent analysis via atomic absorption spectroscopy (AAS) for Fe_carb, Fe_ox and Fe_mag at the University of Leeds. Fe_py was determined gravimetrically following chromous chloride distillation at the University of St Andrew’s. Previous studies have shown reproducibility of better than ±2% (Mettam et al., 2017). Total Fe (Fe_T) was determined after HF–HClO₄–HNO₃ dissolution via AAS. All Fe extractions gave a relative standard deviation (RSD) of < 5 % based on replicate analyses, and sequential extraction analyses were within 5% of the international Fe-speciation reference material WHIT (Alcott et al., 2020). Total dissolution of international sediment standards (USGS; SGR-1bl; USGS SBC-1) gave an Fe recovery of > 98 % .

The sum of Fe_carb, Fe_ox, Fe_mag and Fe_py defines a highly reactive (Fe_HR) pool, which is considered to represent Fe that is biogeochemically reactive during deposition and early diagenesis (Raiswell and Canfield, 1998; Poulton et al., 2004). Anoxic waters commonly have Fe_HR/Fe_T ratios > 0.38, in contrast to oxic depositional conditions where ratios are generally < 0.22 (Poulton and Canfield, 2011). Elevated Fe_HR/Fe_T ratios in anoxic settings arise from the additional water column formation of pyrite in euxinic (sulphidic) settings, or unsulphidized Fe_HR minerals in ferruginous (Fe-containing) settings. Thus, for anoxic samples (i.e., Fe_HR/Fe_T > 0.38) the ratio of Fe_py/Fe_HR distinguishes euxinic (Fe_py/Fe_HR > 0.8) from ferruginous (Fe_py/Fe_FR < 0.6) water column conditions (Poulton and Canfield, 2011; Benkovitz et al., 2020; Poulton, 2021). Fe_HR/Fe_T ratios of 0.22–0.38 are considered equivocal, and may occur due to the masking of water column enrichments via rapid sedimentation (e.g., during turbidite deposition Canfield et al., 1996), or due to transformation of unsulphidized Fe_HR to clay minerals during diagenesis and metamorphism (e.g., Poulton et al., 2010). This second possibility can be evaluated by considering Fe/Al ratios, since Fe_T is preserved even if Fe_HR is lost to clay minerals. In this case, normal oxic marine shales tend to have Fe/Al ratios of 0.55 ± 0.11, and thus Fe/Al > 0.66 is considered to provide a robust indication of water column anoxia (Clarkson et al., 2014).

Trace element composition was determined by ActLabs following their Ultratrace 4 protocol (https://actlabs.com/geochemistry/exploration-geochemistry/4-acid-near-total-digestion/). This is a total digestion using HF, HNO₃ and HCl, followed by analysis via ICP-MS. All measured values are above stated detection limits which are 0.1, 1 and 0.05 ppm for U, V and Mo, respectively. Reported accuracy for all elements are within 1–5% compared to certified standards (GXR-4, SDC-1, DNC-1a, and SBC-1). Reported precisions are ±0.3 ppm for U, ± 2 ppm for V, and ±0.1 ppm for Mo.

Iron speciation data from the Ghaub Formation are characteristic of anoxic, ferruginous bottom water (Figure 2), with Fe_pyrite/Fe_HR ratios below 0.1 and Fe_HR/Fe_total above 0.38. The Mineral fork and Kelly Canyon Formations, however, have Fe speciation data ranging from ferruginous to oxic conditions. In both Utah and Namibia, there appears to be minor fluctuations between anoxic and possibly oxic conditions, but in the samples from the Kelley Canyon Formation reflect oxygenated bottom waters after the Snowball glaciation (Table 2).

In all Utah samples, Mo is low, between detection limit of 0.05 ppm and 0.24 ppm, with the exception of one sample at 4.8 ppm. Vanadium ranges from 20–72 ppm, and U ranges from 0.4 to 6 ppm, with most values above 2.5 ppm (Table 2). These variations are only partially due to differential detrital input, as seen by plotting U normalized to Zr as a function of Al (Figure 3). Zirconium is geochemically similar to U, but Zr is non redox-active, and when plotted against Al (a proxy for detrital input), it is apparent that there is a weak correlation (r ² = 0.36). This indicates that some variation in trace metal abundance may be related to detrital input into the basin, but it is likely not the primary control. In contrast, we observe high U values at low Al concentration in Ghaub samples, indicating authigenic enrichment of U in this location.

There is no clear correlation, however, between the trace element concentration (U and Mo), normalized to Post-Archean Average Shale (PAAS), and total organic carbon (TOC) (Figure 4). By normalizing to PAAS, we are showing whether or not each trace element is enriched compared to the average over time. Such an “enrichment factor”, when plotted as a function of TOC, is an effective redox proxy (Algeo and Liu, 2020). Specifically, this indicates that scavenging by organic carbon during burial is unlikely to be the primary control on trace element abundance. Rather, the abundance of trace elements indicate oxic weathering and delivery of soluble forms of these elements to the water column.

Johnson et al. (2017b) interpreted the Mo, U and V data from Namibia to reflect oxic weathering on the continents, followed by delivery to the ocean and scavenging in ferruginous bottom waters. While U supply is controlled by detrital input, by normalizing to Al (Figure 3), Johnson et al. (2017b) demonstrated that changing U concentrations were not due to changing detrital input alone, and were consistent with iron speciation data. Similarly, these redox-sensitive elements from Utah are present in similar concentrations (Figure 5). Uranium, specifically, also displays a negative correlation with Fe_HR/Fe_Total, consistent with higher U concentrations reflecting a more pervasively oxygenated water column.

There is no enrichment, however, in Mo or V observed from the Mineral Fork or Kelley Canyon formation. Recent work from the early Cambrian of South China has suggested that in settings with a well developed, dynamic redox-cline, the behavior of Mo, V, and U may be decoupled (Han et al., 2018). Specifically, the presence of a zone of Fe-Mn particulate formation can preferentially enrich sediments in V. If such a zone is absent, or sediments sampled are from shallower water, such V enrichment could be missing. It is likely that the Mineral Fork samples were deposited near the ice-grounding line (Christie-Blick, 1982), and perhaps the water column at this location did not contain a Fe-Mn particulate shuttle, leading to low V concentrations.

Overall concentrations of these elements observed from Utah and Namibia are similar to other work focused on the Cryogenian. For example, a section spanning both Sturtian and Marinoan from the Zavkhan Terrane, southwest Mongolia finds U concentrations of between 0.05 and 2 ppm, very similar to our measurements (Lau et al., 2017). Trace element measurements from the Doushantuo Formation of South China, deposited after the Marinoan glacial period, display quite high concentrations of U (up to 30 ppm), V (1,000 ppm), and Mo (200 ppm) in the immediate deglacial period, with values decreasing rapidly up-section (Sahoo et al., 2012). Since the Kelley Canyon formation overlies the cap carbonate, it would have been deposited sometime after the deglaciation, and thus likely does not capture the immediate deglaciation.

We present N concentration and isotope data in Figure 6. δ ¹⁵N in both Namibia and Utah are positive, between about 1 to 5⁰/₀₀ from the Ghaub and 4.1 to 7.6⁰/₀₀ in the Mineral Fork (Figures 6, 7). Values from the Kelley Canyon Formation are higher, between 6.7 to 8.9⁰/₀₀. Nitrogen concentration is similar in units from Namibia and Utah, ranging from 100 to 650 ppm. As noted in Section 3.1, analysis of previously measured till samples from the Timeball Hill and Blaubeker formations produced nearly identical δ ¹⁵N values but different N concentrations. Previous work has noted that within-sample heterogeneity may be an issue for N measurements (Johnson et al., 2017a). Rock hosted N can potentially exist as different species (e.g., NH 4 + , NO 3 − , N₂) and in different organic or mineral phases. So, it is possible that small scale heterogeneity, or slightly different analytical conditions/techniques may be liberating different pools of N. Bulk N isotope values, however, seem to be robust even if measured N concentrations vary (Bräuer and Hahne, 2005; Boocock et al., 2020).

To assess the potential influence of metamorphic alteration of N content and isotopic signals, we show δ ¹⁵N plotted against N concentration, Rb concentration, and C concentration (Figure 6). We observe no correlation between δ ¹⁵N and N concentration, suggesting that higher δ ¹⁵N values are not caused by progressive N loss during metamorphism (Bebout and Fogel, 1992). There is a rough correlation between N and Rb concentration, suggesting some variation in N content could be related to detrital input. The lack of correlation, however, between δ ¹⁵N and N concentration again indicates that while there might be some lithologic control on total N, this does not influence isotope values. There is not a clear correlation between C and N content, which is different than previous work on the non-glacial Neoproterozoic (e.g., Ader et al., 2014), but similar to previous analyses of syn-glacial units (Johnson et al., 2017b). This suggests that there may be either differential preservation of N and C in these units, or that the N and C cycles are differently coupled during Snowball glaciations than the non-glacial intervals of the Cryogenian.

Assuming atmospheric N₂ was at 0⁰/₀₀ during the Cryogenian (Stüeken et al., 2016b), enriched δ ¹⁵N values, given the evidence of persistent O₂ availability provided by redox-sensitive element data, are consistent with the presence of partial, water-column denitrification during the Snowball glaciation. That is, there must have been sufficient O₂ present to support nitrification, with denitrification occurring at boundaries between oxygenated and low-oxygen waters, with a likely chemocline location in bottom waters. The similarity of Mineral Fork δ ¹⁵N values to average modern marine values (5 to 7⁰/₀₀) indicates that a similar relative amount of partial denitrification ( ∼ 20 % ) was occurring in the water column (Tesdal et al., 2013).

After the glaciation, as preserved by the Kelley Canyon formation, samples have the highest δ ¹⁵N values measured. These values are typically interpreted to indicate more extensive denitrification, and lower water-column O₂, to drive residual N to higher δ ¹⁵N values. Yet, Fe-speciation data indicates oxygenated bottom waters during the deposition of the Kelley Canyon Formation (Figures 2, 7). Kelley Canyon units also have the highest U concentrations of any samples presented (Figure 5), suggesting oxic weathering and delivery of U to the site of deposition. Thus, we require an interpretation that can explain both higher δ ¹⁵N values and a pervasively oxygenated water column. We propose two potential explanations: oceanographic circulation shifts and biologic responses to deglaciation.

A similar pattern, positive but low and then increasing several permil, in δ ¹⁵N values in sediments has been observed in a number of locations during the Pleistocene-Holocene deglaciation (Galbraith and Kienast, 2013). Several sites are near-shore environments, including one off the coast of northern Chile (De Pol-Holz et al., 2006). De Pol-Holz et al. (2006) observed an increase in sedimentary δ ¹⁵N values from 8 to 12⁰/₀₀ across the deglaciation, and attributed this increase to melting of the Patagonian Ice Sheet. The influx of freshwater slowed formation of intermediate water, which in turn decreased deep water ventilation. This oceanographic reorganization could have resulted in upwelled N that was enriched in ¹⁵N, due to removal by denitrification, compared to the glacial maximum with a less well developed oxygen minimum zone. The Snowball deglaciation would have produced a large meltwater lid (Hoffman PF. et al., 2017), which could affect ocean circulation in a similar way: limited deep-water ventilation, increased denitrification at a chemocline underlying a well oxygenated upper water column.

In the wake of both the Sturtian and the Marinoan glaciations, the ocean appears to have become more oxygen rich and the surface temperature increased (Hoffman PF. et al., 2017; Lechte et al., 2019). There are many studies on contemporary denitrifying bacteria, both in the field and in culture, that demonstrate an increase in denitrification activity at higher temperatures (Nowicki, 1994; Veraart et al., 2011). Thus, we would expect to observe an increase in δ ¹⁵N in post-Snowball warm oceans, observed in both the Ghaub and Mineral Fork-Kelley Canyon Formations. While an increase in water column oxygen may seem anathema to increased denitrification, modern denitrifiers do function in oxygenated water. Specifically, by using the Nap reductase enzyme, instead of the Nar enzyme (Berks et al., 2001), organisms can perform aerobic denitrification (Ji et al., 2015). Perhaps, then, the combined effect of high temperatures and aerobic denitrification caused an increase in δ ¹⁵N values in the post-Snowball ocean, even in the presence of higher water column O₂.

Alternately, it is possible that enriched δ ¹⁵N values could reflect partial nitrification followed by complete denitrification. This scenario was proposed to explain very positive δ ¹⁵N values from the 2.7 Ga Tumbiana Formation (Thomazo et al., 2011), though an alternate explanation has been proposed suggesting high values could be caused by alkaline conditions and degassing of NH₃ (Stüeken et al., 2015). Nitrification to nitrite has a pronounced isotopic effect, with product NO 2 − depleted between 13 to 38⁰/₀₀ compared to reactant NH 4 + (Casciotti et al., 2003). If, then, NO 2 − is quantitatively removed via anammox, the residual N pool would be comprised of isotopically enriched NH 4 + . In the modern ocean, this scenario has been observed in sediments of the Bering Sea (e.g., Morales et al., 2014). There is, however, a much wider range in sediment δ ¹⁵N values (+2⁰/₀₀ to + 18⁰/₀₀), reflecting the transient nature of the geochemical conditions necessary for this to occur. Since the N isotope data from both Cryogenian sections are more tightly coupled with evidence for persistent oxic weathering, we suggest that they instead reflect less transient redox conditions, and likely complete nitrification followed by partial denitrification.

Additionally, we note that δ ¹⁵N values are generally lower in samples with Fe_HR/Fe_total values that are ‘equivocal’ compared to those that clearly plot in the ferruginous field (Figure 2, 7). Since this potential trend is not strongly correlated (r ² = 0.13), any interpretations at this point are speculative. It is possible that samples with low Fe_total, and therefore higher Fe_HR/Fe_total, represent upwelling into oxygenated conditions, driving denitrification and increasing δ ¹⁵N values. However, iron speciation data does not allow for quantitative interpretation of bottom water conditions within fields, but rather serves to distinguish between overarching redox conditions. Future work, either observational or geochemical modeling, could be valuable to investigate any controls on patterns within Fe-speciation fields.