Persistently low atmospheric oxygen requires that net organic carbon burial was muted through much of Earth’s middle age. In order to achieve global mass balance with respect to O2, recent models have suggested that redox-dependent mechanisms, such as Fe(II)-phosphate precipitation, limited phosphate availability in dominantly anoxic and ferruginous oceans, in turn limiting net primary production, and therefore organic carbon burial. Nevertheless, observational constraints on phosphorus cycling in ferruginous Proterozoic systems are rare, leaving these models largely untested. Here, we present high-resolution petrographic and mineralogical data showing that the 1.3 Ga Sherwin Ironstone (Roper Group, Australia) was dominated by syndepositional precipitation of the Fe(II)-silicate minerals greenalite and berthierine, interlaminated with abundant authigenic calcium fluorapatite (CFA). Set in a quantitative geochemical framework, these data reveal that elevated marine SiO2(aq) concentrations facilitated extensive Fe(II)-silicate production, leaving CFA, rather than Fe(II)-phosphate, as the principal inorganic phosphorous sink in shallow-water Roper Group sediments. More broadly, the physical and chemical factors that triggered Fe(II)-silicate and CFA burial in the Roper Seaway highlight semi-restricted basins as important loci of phosphorus removal from the mid-Proterozoic ocean.

Low atmospheric oxygen is considered a defining characteristic of the Precambrian Earth, and recent models have highlighted the biogeochemical implications of a stable low-oxygen world. In order to stabilize low atmospheric O2, some models hypothesize that the rate of net primary production (NPP) must have been lower (up to 10×–100×) than at present, reducing organic carbon burial (Derry, 2015; Laakso and Schrag, 2018a). Phosphorus limitation has emerged as the most likely mechanism to have attenuated NPP (Laakso and Schrag, 2018a, 2018b) and organic carbon burial. Several hypotheses have focused on redox-dependent mechanisms that might have limited P availability in anoxic and ferruginous oceans (Laakso and Schrag, 2014; Derry, 2015). Recently, a compilation of the total phosphorous concentration ([P]) in shales has revealed a four-fold increase that began in the early to mid-Neoproterozoic, continuing to modern times (Reinhard et al., 2017). This record is thought to reflect a major shift from redox-dependent P limitation to widespread P availability in the water column with the demise of Fe-rich waters (Reinhard et al., 2017).

However, direct observations that constrain P cycling in ferruginous Proterozoic systems are rare. For example, low [P] in many mid-Proterozoic shales has been interpreted to reflect P limitation through Fe(II)-phosphate precipitation in ferruginous systems (Derry, 2015), P scavenging through green rust precipitation (Zegeye et al., 2012), and adsorption and /or coprecipitation with Fe(III)-oxides (Jones et al., 2015). Although these mechanisms receive support from environmental analogues and geochemical models, they have not been tested against detailed mineralogical and sedimentological data from Proterozoic rocks. In addition, the stratigraphic distribution of syndepositional and early diagenetic chert in Proterozoic rocks reflects a marine system characterized by elevated SiO2(aq) concentrations (Maliva et al., 2005), yet silica does not feature in current models or analogues for the mid-Proterozoic P cycle.

Here, to investigate the role of ferruginous and SiO2-rich conditions on phosphate cycling in shallow-marine environments, we conducted a high-resolution sedimentological, mineralogical, and geochemical study of the 1.3 Ga Sherwin Ironstone of the Roper Superbasin, Northern Territory, Australia. The Sherwin Ironstone records an episode of Fe-rich chemical sedimentation in the Mesoproterozoic, offering a valuable window into contemporaneous Fe and P cycling.

The Roper Superbasin hosts an exceptionally well-preserved record of sedimentation across multiple epicratonic basins of northern Australia (Munson, 2016). Sediment fill is packaged into six transgressive sequences of basinal shales and proximal sands, bound by intercycle erosional surfaces and ironstones (Munson, 2016). Local flexural tectonics controlled both relative sea level and sedimentation, whereby erosional loading of the basin caused subsidence in the depocenter (Abbott and Sweet, 2000; Munson, 2016).

Iron-rich chemical sediments are common across the Roper Superbasin as sequence-bounding ironstones and/or sandstones with Fe-rich cement, which were deposited during the initial stages of regional transgression (Abbott and Sweet, 2000). The most significant of these, the 1.3 Ga Sherwin Ironstone (Fig. 1; Yang et al., 2018), is dominated by iron-cemented sandstones and subordinate beds of Fe-silicate ooids and laminated Fe-silicate chemical sediment, reflecting shallow-marine upper-shoreface sedimentation (Cochrane and Edwards, 1960; Munson, 2016).

To examine chemical sedimentation associated with the Sherwin Ironstone, we logged and sampled six drill cores through the unit and examined 120 thin sections by optical microscopy and scanning electron microscopy–energy dispersive X-ray spectroscopy (SEM-EDS) using an FEI Quanta 650 FEG SEM. Subsamples of petrographic facies were microdrilled and analyzed for bulk and clay-specific mineralogy at the University of Oxford (Oxford, UK) using a Panalytical Empyrean Series 2 X-ray diffractometer (XRD) equipped with a Co Kα source at 40 kV and 40 mA (see the GSA Data Repository1).

Petrographic and mineralogical data (Figs. 2 and 3) delineate three facies of chemical sediments: (1) laminated Fe-silicate–rich beds, (2) beds of Fe-silicate ooids with erosional surfaces and high-angle cross bedding, and (3) medium- to coarse-grained quartz sand cemented by various Fe minerals (Cochrane and Edwards, 1960). In all facies, authigenic berthierine and greenalite represent the principal building blocks of the Sherwin Ironstone (Fig. 2). The Sherwin Ironstone was deposited in a range of high-energy shallow-water settings, from shoaling bars to upper shoreface, as indicated by abundant shallow-water sedimentary features (Cochrane and Edwards, 1960; Munson, 2016). Recent experimental work has shown that greenalite precipitation occurs under fully anoxic and ferruginous conditions (Tosca et al., 2016). Unlike berthierine (an Fe(II)- and Al-bearing silicate), the formation of which requires pore-water Fe2+ and preexisting aluminosilicate minerals (Bhattacharyya, 1983), the near absence of Al in the greenalite structure facilitates nucleation from the water column (Tosca et al., 2016; Fig. DR7 in the Data Repository). Thus, ferruginous waters commonly expanded onto the shallow shelf in the Roper Superbasin.

Powder XRD and SEM-EDS (Figs. 2 and 3) indicate that calcium fluorapatite (CFA) is an important constituent of the Sherwin Ironstone. CFA occurs as aggregates of anhedral grains within authigenic Fe-silicate cements, as well as within Fe-silicate ooid laminae (Figs. 2 and 3). In both cases, textural relationships indicate that CFA precipitated along with Fe-silicates (Fig. 3). CFA within Fe-silicate ooids (Fig. 3) indents underlying Fe-silicate lamina, indicating that the ooid was plastic when the CFA adhered to its outer surface. Further deformation probably occurred during reworking of the ooids, as indicated by their spastolithic texture. These petrographic relationships indicate that both Fe-silicates and CFA are primary precipitates. Detrital apatite in the Sherwin Ironstone can be distinguished based on its tabular habit, enrichment in Ca and rare earth elements, and lack of detectable F. Our petrographic observations also indicate that most Fe-oxides present within the Sherwin Ironstone are the products of late-stage alteration (Fig. 3B; Cochrane and Edwards, 1960; Data Repository).

Together, petrographic and mineralogical constraints indicate that CFA was deposited with Fe(II)-silicates of the Sherwin Ironstone at or just below the seafloor (Figs. 3A, 3C, and 3D; Fig. DR6). Although these data do not rule out the possibility that some of the PO4 in CFA minerals was derived through dissolution of Fe-oxides, they support the inference that the high aqueous PO4 concentrations necessary for CFA nucleation extended above the sediment-water interface. In addition, the decoupling between detrital mineral phases and CFA suggests that aqueous PO4 accumulation was not strongly linked to input derived from riverine sources (Fig. 2). Similarly, while organic-matter diagenesis would be expected to have released PO4 and fuel CFA precipitation (i.e., an expression of a process known as “sink-switching”; Ruttenberg and Berner, 1993), the Sherwin Ironstone is conspicuously organic poor, with total organic carbon <0.1 wt% (data not shown). Relatively high PO4 concentrations in the water column could also be due to recycling of organic matter derived from primary production (i.e., Lenton and Daines, 2018; Poulton, 2017) or a limited supply of fixed nitrogen (i.e., Sanchez-Baracaldo et al., 2014; Koehler et al., 2017).

Regardless of how PO4 accumulated in the water column during Sherwin Ironstone deposition, our data demonstrate that CFA was an important terminal P sink during Roper Group chemical sedimentation, precipitating contemporaneously with Fe(II)-silicates from a ferruginous water column. Available constraints on apatite and greenalite formation pathways define the chemical solution space required to stabilize these two minerals in Roper Group sediments. Once a specific combination of pH, [SiO2], and [Fe2+] is crossed, then an Fe(II)-silicate gel precipitates, and eventually reorders to form greenalite (Tosca et al., 2016). Assuming saturation with respect to amorphous silica (supported by shallow-water chert deposition in proximally equivalent sediments), minimum [Fe2+] at greenalite nucleation may be determined across a range in pH from 7.2 [below which Fe(II)-silicate precipitation is too slow to occur over reasonable diagenetic time scales; Tosca et al., 2016] to 8.2, a reasonable upper limit based on current constraints on Proterozoic pH (Halevy and Bachan, 2017). Both experimental and observational evidence support an apatite formation pathway that first involves the nucleation of octacalcium phosphate, then subsequent recrystallization to apatite (Jahnke et al., 1982; Gunnars et al., 2004; Van Cappellen and Berner, 1991). Thus, a conservative upper limit for total [PO4] may be determined across the same pH window if marine [Ca2+] is known. Although few constraints on Mesoproterozoic marine [Ca2+] are available, 2.5 mmol/kg serves as a reasonable minimum, below which Ca-sulfate would not precipitate upon evaporation (inconsistent with Roper Superbasin and other Mesoproterozoic sediments; Grotzinger and Kasting, 1993). An upper limit of 30 mmol/kg [Ca2+] serves as a maximum based on Ca-isotope constraints of marine evaporites of the ca. 2.0 Ga Tulomozero Formation (northwestern Russia), and from evaporite geochemistry of the ca. 1200 Ma Society Cliffs Formation (northern Baffin Island, Canada) which indicates that [Ca2+], [K+], and [Na+] were close to modern values (Kah et al., 2001; Blättler et al., 2018). Combining these constraints yields the saturation state of Roper Superbasin water relative to vivianite, a common Fe(II)-phosphate mineral known to precipitate in modern marine and lacustrine sediments (more readily than the Mg-rich end members bobierrite and baricite; Egger et al., 2015; Dijkstra et al., 2016; Fig. 4). These calculations show that for all conditions except those associated with alkaline pH and very high marine [Ca2+], the contemporaneous precipitation of greenalite and Ca-phosphate occurred despite significant vivianite supersaturation.

This combination of petrographic and mineralogical data with geochemical constraints reveals a significant kinetic barrier associated with vivianite precipitation, a conclusion reinforced by pore-water data from modern sediments. For example, a number of studies have reported significant supersaturation with respect to Fe(II)-phosphate in lacustrine, estuarine, and marine systems, despite the presence or absence of solid Fe(II)-phosphate phases (Postma, 1981; Rothe et al., 2016). This further indicates that when the threshold for nucleation has been reached in some diagenetic systems, Fe(II)-phosphate minerals may not grow rapidly enough to buffer PO4 to equilibrium solubility levels. This kinetic limitation implies that Fe(II)-phosphate precipitation should occur in sedimentary environments that are routinely associated with excessive supersaturation. In fact, a number of recent studies have shown that modern marine Fe(II)-phosphate precipitation is dominantly confined to sedimentary environments below the sulfate-methane transition zone (SMTZ; Egger et al., 2015; Dijkstra et al., 2016; März et al., 2018).

One significant difference between modern sedimentary environments conducive to Fe(II)-phosphate nucleation and the Sherwin Ironstone is the absence of other Fe(II)-bearing phases that kinetically compete for Fe2+. Indeed, at the Fe2+ concentrations and alkalinities associated with modern sediments below the SMTZ, very low SiO2(aq) concentration typically precludes Fe(II)-silicate formation. However, in Proterozoic marine systems where SiO2(aq) was much higher than in the modern ocean (Maliva et al., 2005), experimental constraints indicate that high [Fe2+] and alkalinities [and thus Fe(II)-phosphate supersaturation] may not have been reached without rapid nucleation of Fe(II)-silicates such as greenalite (Tosca et a., 2016). As our data indicate, the presence of elevated SiO2(aq) in mid-Proterozoic marine systems appears to have efficiently sequestered Fe2+ in Fe(II)-silicate phases, while leaving CFA as the dominant inorganic sink for phosphorus.

Our conclusions receive additional support from a number of examples where Fe(II)-silicate minerals have precipitated in close association with Ca-phosphate; for instance, the Paleoproterozoic Sokoman Iron Formation of the Labrador Trough, eastern Canada (Pufahl et al., 2014), and Ordovician oolitic ironstones of Bell Island, Newfoundland, Canada (Todd et al., 2018). Oolitic ironstones composed of Fe-silicate and authigenic phosphates also occur in the early Jurassic Cleveland Ironstone, northern England, UK, in the Passwang Formation, Switzerland (Burkhalter, 1995; MacQuaker and Taylor, 1996), and in the Cretaceous of southeastern Australia (Boyd et al., 2004).

Together, a variety of data indicate that elevated [SiO2] and Fe(II)-silicate precipitation may have prevented Fe(II)-phosphate burial from Proterozoic seas, in contrast with models that envisage a link between PO4 limitation and ferruginous chemistry (Planavsky et al., 2011; Derry, 2015; Reinhard et al., 2017). Nevertheless, the requirement for global oxygen mass balance implies that other mechanisms may have regulated marine PO4 availability through the Proterozoic (Diaz et al., 2008; Cox et al., 2019). Our analyses also indicate that shallow epeiric seas such as the Roper Superbasin served as key depositional loci for both Fe(II)-silicates and Ca-phosphates. Unique physical and chemical factors, such as partial regional restriction coupled with enhanced alkalinity input from surrounding volcanic catchment (Cox et al., 2016), established these environments as permanent chemical sedimentary traps for both riverine and deep-marine (i.e., upwelling-derived) PO4. This unusual set of geological conditions, apparently common in the mid-Proterozoic, would have impacted the global marine PO4 reservoir until the Tonian–Cryogenian breakup of Rodinia (Horton, 2015). A tectono-sedimentary influence on P burial also reconciles shale-hosted P and Fe-speciation records, which, when viewed together, suggest that Proterozoic P burial increased even though anoxic and/or ferruginous waters persisted well into the Paleozoic (Sperling et al., 2015). Nevertheless, further examination of Fe and P mineralogy preserved in ancient sediments will continue to reshape our understanding of Proterozoic biogeochemical cycling.

We are grateful to Amber Jarret and Grant Cox for facilitating sample acquisition and helpful discussions. Staff at Geoscience Australia and the Northern Territory Geological Survey are thanked for provision of sample material and generous assistance with sample collection. Katherine Clayton collected the XRD data. Fatima Ali and Tobermory Mackay-Champion assisted with sample preparation and data collection, and Jon Wade assisted with acquiring SEM-EDS data. Steve Wyatt and Phil Holdship assisted with ICP-MS analysis. BRJ was funded by Natural Environment Research Council award NE/L002612/1. We thank Malcolm Wallace and two anonymous reviewers, whose thoughtful comments improved this manuscript.

1GSA Data Repository item 2020031, age models, lithofacies details, extended methods, detailed discussion of vivianite formation, and additional data figures, is available online at, or on request from
Gold Open Access: This paper is published under the terms of the CC-BY license.