Superheavy S isotopes from glacier-associated sediments of the Neoproterozoic of south China: Oceanic anoxia or sulfate limitation?
Published:January 01, 2006
- PDF LinkChapter PDF
Liu Tie-bing, J. Barry Maynard, John Alten, 2006. "Superheavy S isotopes from glacier-associated sediments of the Neoproterozoic of south China: Oceanic anoxia or sulfate limitation?", Evolution of Early Earth's Atmosphere, Hydrosphere, and Biosphere - Constraints from Ore Deposits, Stephen E. Kesler, Hiroshi Ohmoto
Download citation file:
Black shales and Mn carbonates interbedded with glacial deposits from the Neoproterozoic of southern China exhibit extremely heavy values of pyrite S isotopes that may reflect the peculiar environment of Earth at this time. δ34S averages +30‰ at Tanganshan and +44‰ at Xiangtan, compared with typical values of 0‰ to +5‰ found in younger deposits. Furthermore there is no distinction between the shales and the Mn carbonate ores in the Neoproterozoic, unlike the younger deposits, which show much lighter δ34S in the shales than in the Mn ores (the spread is 25‰). Most other chemical parameters are very similar to both the younger Mn deposits and those from the Paleoproterozoic. The exception is the rare earth elements (REE). All Neoproterozoic Fe ores and most Neoproterozoic Mn ores lack the positive Eu anomaly that characterizes Archean and Paleoproterozoic Fe-Mn accumulations. On the other hand, Neoproterozoic Mn deposits have positive Ce anomalies on North American Shale Composite (NASC) normalized plots, in contrast to other MnCO3 ores. The ΣREE is also higher than in other Mn deposits, but lower than in modern deep-sea crusts.
Sulfide S values in all Neoproterozoic shales tend to be exceptionally variable and to often show much heavier values than can be found in marine strata from the Phanerozoic. Therefore the anomalous δ34S values we observed reflect peculiar conditions in the world oceans at this time rather than purely local effects. Times of enrichment of seawater sulfate in 34S do not correspond to periods of glaciation, so the likely cause of the S isotope patterns is not worldwide glaciation, but a generally low level of dissolved sulfate S in the Neoproterozoic oceans that allowed modest increases in the amounts of S removed as pyrite to drive down the oceanic S reservoir enough to produce strong Rayleigh reservoir effects. The abundance worldwide of Sturtian-age Mn and Fe deposits indicates an increase in Fe flux to the oceans that would have been sufficient to depress SO4 2- levels severely and to result in residual dissolved S extremely enriched in 34S. REE evidence indicates that most of this enhanced Fe and Mn flux came from diagenetic remobilization of detrital oxides rather than from ridge-crest hydrothermal systems, in contrast to the Paleoproterozoic banded iron formations. Rapid introduction of lateritic soil residues to restricted basins by low-latitude glaciation could have provided the needed excess Fe and Mn to drive this system.
The Neoproterozoic was a time of profound change in Earth's oceans and biosphere. One peculiar manifestation of these changes is the reappearance of iron formations in abundance in the sedimentary record, for example Rapitan in Canada (Yeo, 1981), the iron-rich beds of the Chuos Formation in Namibia (Breitkopf, 1988), and the Braemar Ironstone of South Australia (Lottermoser and Ashley, 2000). Accompanying these iron formations are some of the world's most important Mn ore deposits, e.g., Urucum in Brazil (Urban et al. 1992), and in all cases the ore bodies are found in close association with glacial deposits (Maynard, 1991). In a study of the relationship of Mn ore genesis to black shale sedimentation, Liu (1988, 1990) compared the whole-rock carbon-sulfur geochemistry of one of these Neoproterozoic deposits, Xiangtan, with younger deposits and found the patterns to be similar. Subsequent work on the sulfur isotopic composition of the Chinese Neoproterozoic deposits, however, has shown them to contain extraordinarily heavy S (Tang and Liu, 1999; Li et al., 1999).
Is this super-heavy sulfur peculiar to the Neoproterozoic of south China? Is it somehow related to the Mn-Fe mineralization? Is it somehow related to glaciation? To address these questions we have undertaken detailed geochemical analysis of two of these deposits, Xiangtan and Tanganshan. In this paper we show how isotopic, trace element, and rare earth element (REE) data constrain models for the genesis of Mn-Fe deposits of this time period and help understand the Neoproterozoic atmosphere-ocean system.
There are 11 large to medium-sized commercial Mn operations in Neoproterozoic host rocks of south China distributed throughout the Hubei, Guizhou, Sichuan, and Hunan Provinces (Xu et al., 1990; Fan et al., 1992; Li et al., 1999). We studied Xiangtan and Tanganshan in Hunan province (Fig. 1). The ores are contained within black shales that are overlain and underlain by glacial deposits (Fig. 2). The regional stratigraphic section comprises, from the base, the Liantuo Formation sandstones, the Gucheng or Chunmu sandstones and diamictites, the Datangpo Formation (Minle Formation of some authors), comprising shales and Mn carbonates, and, at the top, the Nantuo Formation diamictites. The Nantuo is overlain abruptly by carbonates of the Doushantuo Formation and this contact is an important stratigraphic marker in south China. Thus there are two diamictite intervals separated by a black shale with Mn carbonate beds.
Ages are sparse, but a tuff bed in the Liantuo Formation has yielded one date of 748 ± 12 Ma from single-zircon U-Pb dating, and the Datangpo/Minle Formation has been dated by Rb-Sr at 728 ± 27 Ma (Li et al., 2000). These ages and sequence stratigraphic interpretations led Wang and Li (2003, p. 155) to suggest that the Nantuo diamictites are correlative with the Sturtian glacial deposits of Australia. If so, the Datangpo Mn mineralization is comparable to the Braemar ironstone, which is interbedded with the glacial deposits (Gorjan et al., 2000). The overlying Doushantuo carbonates, however, have been dated to 584 ± 26 Ma by Lu-Hf and 599.3 ± 4.2 Ma by Pb-Pb methods (Barfod et al., 2002), which suggests that the Nantuo glacial deposits correspond to the later Marinoan glaciation episode. Dobrzinski et al. (2004), however, argued that both the Chunmu and the Nantuo diamictites are Sturtian, on the basis of the 728 Ma date for the intervening Datangpo Formation. Paleomagnetic data are also ambiguous. Rui and Piper (1997) assigned the Datangpo Formation to the Sturtian, whereas Macouin et al. (2004) favored a Marinoan age. Both Shen (2002) and Macouin et al. (2004) argued for a Marinoan age based on carbon isotope excursions, but either major glacial episode could produce this effect. Thus there are two conflicting age interpretations. If the Marinoan age is correct, then perhaps the Chunmu tillite is a Sturtian equivalent and the Datangpo Formation represents a relatively long time period. Alternatively, the Chunmu and Nantuo tillites could both be phases of Sturtian glaciation and the Datangpo was rapidly deposited during a brief interglacial episode.
The reconstruction of Xu et al. (1990) shows the Datangpo Formation lying within a fault-bounded marine trough between the Yangtze Block to the northwest and the Hunan Block to the southeast. Continental diamictites occupied the margin of the Yangtze Block during Datangpo deposition, whereas glaciomarine strata are found on the Hunan Block. Thus deposition was in a narrow fjord-like rift valley. Urban et al. (1992) proposed a similar ice-covered long, narrow basin as the setting for the Urucum deposits of Brazil. Eyles and Januszczak (2004a, 2004b) have likewise emphasized the rift setting of Neoproterozoic glacial deposits as key to their genesis.
Samples were collected from mine passageways and were ground in a tungsten carbide ring mill. They were analyzed for major elements by X-ray fluorescence (XRF) and for trace elements by inductively coupled plasma–mass spectrometry (ICP-MS) supplemented by hydride-atomic absorption (AA) for As and Se and by LECO elemental analyzer for C and S. XRF analyses were done at the University of Cincinnati using a Rigaku 3070 spectrometer. Loss-on-ignition residues of the samples were fused with lithium metaborate. The resulting glass beads were then ground and pressed into disks with polyvinyl alcohol binder. Calibration was done by empirical multiple regression against a set of U.S. Geological Survey and National Institute of Standards and Technology (NIST) high-Mn standards treated in the same way. Reproducibility is ±2%. ICP-MS and AA were performed by XRAL Laboratories of Toronto, Canada. Precision is about ±5% for most elements with detection limits generally of 0.1–1 ppm for trace elements and 0.05–0.1 for REE. LECO analyses were performed on whole-rock samples and on acid-insoluble residues at Indiana University. Precision was ±1% for C and ±5% for S.
Sulfur was extracted from the samples by a method originally developed by Canfield et al. (1986) and modified by the biogeochemistry laboratory at Indiana University (Bruchert, 1998). First, sample powders are extracted for 24 h with methylene chloride in a soxhlet apparatus. This procedure removes native sulfur and bitumen-bound sulfur. Extracting the bitumen also removes organic coatings from sulfide mineral grains and makes them easier to dissolve in subsequent steps. Any extracted native sulfur is reprecipitated as copper sulfide on copper shot in a receiving flask. The S-coated copper shot is then transferred to a closed reaction system where it is treated with 6N HCl under a nitrogen stream and heated to 90 °C. This process converts the copper sulfide to H2S, which is bubbled into a AgNO3 trap where the S is precipitated as Ag2S. Next 0.5–2.0 g of the MeCl-extracted samples are transferred to the same reaction system. Heating the sample to 90 °C in 6N HCl releases S contained in monosulfides (FeS, galena, sphalerite) and also in pyrrhotite. This sulfide converts to H2S in the acid environment, which is again bubbled into a AgNO3 trap. This S is customarily designated as the acid-volatile fraction (AVS). The sulfur in pyrite and marcasite, FeS2, occurs in the S1- form and must be reduced to the S2- form to be converted to H2S gas. This reduction is accomplished using chromous chloride in 6N HCl, again at 90 °C, followed by precipitation of Ag2S from the H2S gas. We designate this S as the chrome-reducible fraction (CrRS). The Xiangtan and Tanganshan samples yielded only pyrite S, except for two samples from the ore dump at Tanganshan that had large amounts of native S. We interpret this native S as a weathering product of pyrite and have not investigated surface samples further because of the large impact of weathering, even with relatively brief periods of exposure. Sulfur isotopic ratios were determined at Indiana University using the Finnigan MAT 252 mass spectrometer described by Studley et al. (2002).
Whole Rock Chemistry
The Chinese Mn ores have extreme S isotope values, and might be expected to also have significant anomalies in other chemical properties. However, whole rock data for the Mn ores and for the host black shales are similar to values reported from other Mn deposits and for representative unmineralized black shales. In Table 1 we give major element results for the Xiangtan and Tanganshan ores and shales plus analytical data using the same procedures for a set of NIST standards of Mn and Fe-rich rocks and a representative suite of ore samples from Paleoproterozoic and Phanerozoic Mn deposits. Table 2 gives trace element results and Table 3 gives REE results from the same samples and standards. Many of the NIST standards do not have trace and rare earth elements included in their documentation, so we present values for these standards for the first time. Mamatwan is the largest Paleoproterozoic deposit and comprises a significant fraction of the world's Mn reserves. Molango is the largest Mesozoic deposit, and Taojiang, although much smaller, is representative of Paleozoic deposits. For each deposit, we have selected representative ore samples, either Mn-rich or Fe-rich, and some host rocks. Extensive trace element and REE data for these deposits have not previously been published.
Inspection of the tabulated results for the ores and shales in the Chinese deposits shows a continuum of compositions between high and low Al end members. The data set shows a strong positive correlation between % Mn and % Ccarbonate (r2 = 0.74) and a negative correlation between % Mn and both % Corg (r 2 = −0.77) and %Al2O3 (r 2 = −0.71). No other elements show a positive correlation to % Mn and there are no correlations between any of the whole rock parameters and δ34S.
Thus the highest-grade ore samples from the Chinese deposits are low in Al and Corg, but high in Ccarb. We have used 3% Al2O3 as the upper cutoff to separate ore samples from subore grade, but mineralized, samples. In Figure 3 we compare the composition of this end-member ore type to an average MnCO3 ore as calculated from the representative ores given in Tables 1 and 2, giving equal weight to ores of each of the three ages. Two groups of elements show enrichment in the Chinese ores: one group comprising P and Mo, which are probably related to the high organic carbon content of the ores, and another comprising Ti, Y, Zr, and Nb, which are elements typically associated with heavy minerals. Al is identical in the two data sets, so these excess resistate elements are not associated with clays. Pb is also relatively high in the Chinese ores and Ca and Ta are depleted.
Because the Chinese ores are hosted by black shales, we have used the geostandard SDO-1 as a comparison. The Chinese black shales are very similar to this standard (Fig. 4). Compared with SDO-1, Mn is somewhat enriched in the Chinese host shales, so Mn mineralization is distributed at low grade in the host rocks, a pattern that is also seen at Molango. Mo and U are low compared to SDO-1, which is probably related to much higher organic carbon in the standard (10% compared to ∼3%–4% for the Chinese shales). Thus, just as for the ores, the host rocks appear to have normal chemistries except for their S isotopic values.
For the REEs, some differences can be seen between the Chinese material and other ores (Fig. 5). Almost all of the older and younger Mn deposits in Table 3 have negative Ce anomalies. Both the Chinese ore deposits by contrast display positive anomalies. Yang et al. (1999) also reported a positive Ce anomaly for one sample from the correlative Songtao deposit. For Eu, the samples from other deposits all have positive Eu anomalies, normalized to NASC. For the Chinese deposits, Tanganshan also has a positive anomaly, whereas all of the samples from the Xiangtan deposit have distinct negative Eu anomalies. The La/Yb ratio is less than 1.0 for both of the Chinese deposits, whereas the ratio is substantially greater than 1.0 for the others. On the other hand, the ΣREE is considerably higher in the Neoproterozoic deposits. These differences in REE chemistry are not related to detrital contamination by heavy minerals; there is no correlation between ΣREE and either P2O5 or Zr content of the ores, so xenotime or zircon are not controlling the REE chemistry of the ores. Therefore the signature is mostly carried by the Mn minerals themselves and is a fundamental character of the deposits.
In studies of black shales like the host rocks to these deposits, a common question is the degree of anoxia in the basin during deposition. Several ways to estimate this property have been proposed. Among the most common are degree of pyritization (Raiswell et al., 1988), C-S plots, and Mo content (Crusius et al., 1996). Liu (1988) measured degree of pyritization (DOP) in the shales hosting a number of Mn deposits and reported for Xiangtan that footwall shales average 0.70, the ores average 0.49, and the hanging wall shales average 0.71. For comparison, Raiswell et al. (1988) suggested 0.45 as the boundary between fully oxic and suboxic environments and 0.75 as the boundary between suboxic and fully anoxic. The Xiangtan ores, using this scale, were deposited under oxic to suboxic conditions, whereas their host shales were deposited under conditions transitional between suboxic and anoxic. Comparable data are not available for Tanganshan.
In Figure 6 we show a plot of organic C versus total S for Xiangtan and Tanganshan. The data show a pattern typical of anoxic-euxinic conditions, in which there is a poor correlation between S and Corg, with S falling mostly between 2% and 4%. Thus the C-S relations indicate more reducing conditions than does DOP.
Mo/Al ratios are commonly used to assess the extent of Mo enrichment in shales (e.g., Warning and Brumsack, 2000). For the Chinese shale samples, the Mo/Al (ppm metal/% metal) ratio at Tanganshan averages 12, whereas at Xiangtan it averages only 3. For comparison, the ratio is to 16–20 for modern euxinic sediments and <1 for average shale. A comparison of Mo versus Corg (Fig. 7) shows no correlation, contrary to the situation seen in most modern sediments, which show an excellent correlation. Furthermore, the Mo contents are almost all much lower than for modern euxinic sediments. The DOP and C/S patterns suggest that the host shales for the Chinese Mn deposits were formed under transitional suboxic-anoxic conditions and that the degree of anoxia was greater at Tanganshan than at Xiangtan. However, Mo levels are somewhat lower in these shales than we would expect from the pattern seen in modern sediments. Euxinic basins sequester large amounts of Mo as MoS2 in bottom sediments (Barling and Anbar, 2004). Perhaps the low Mo in these Neoproterozoic black shales is related to widespread oceanic anoxia, with a higher flux of Mo to deep-water sediments than today, which resulted in less Mo available for near-shore deposition. Dobrzinski et al. (2004) have performed a similar paleoredox analysis using S/C, U/Th, Cd, Mo, and Ce* as indicators. Their focus was largely on the glacial marine units, which they assigned to oxic depositional conditions. The interglacial interval, however, was more reducing, falling at the border between oxic and suboxic or between suboxic and anoxic, depending on the indicator.
The Chinese Neoproterozoic Mn ores are radically different in S isotope chemistry from other Mn ores. Figure 8 shows the distribution of S isotopes in MnCO3 ores and host black shales from Phanerozoic deposits. Note that the Phanerozoic shales yield broadly negative δ34S values with a mode at about −25‰. This distribution is typical of syngenetic pyrite formed by bacterial sulfate reduction in fine-grained sediments (e.g., see Ohmoto and Goldhaber, 1997, Fig. 11.17). The MnCO3 ores, on the other hand, have a more irregular distribution with a distinctly higher mode at about +5‰. Okita (1992) explained this heavier S as resulting from formation of pyrite deeper in the sediment than is normally the case during the diagenesis of marine sediments. The large volume of Mn in these sediments poises the Eh at values too high for sulfate reduction. Until all Mn oxide is converted to Mn carbonate, Eh stays high and no sulfides form. Thus pyrite formation takes place farther from the sediment-water interface than usual. This greater depth reduces diffusive exchange with the bulk seawater, resulting in complete reduction of a smaller sulfate reservoir and thus smaller amounts of pyrite with heavier δ34S than normal.
In Figure 9 we show our data plus previous results from similar Chinese deposits reported by Li et al. (1999) for the Xiangtan and Songtao deposits and by Tang and Liu (1999) for the Minle deposit. The Neoproterozoic ores and their host shales are uniformly shifted to very enriched δ34S values, with no distinction between the ores and the host shales. Xiangtan and Tanganshan deposits have significantly different δ34S, although both are far more enriched in 34S than other deposits shown in Table 1. The average for Xiangtan is +44.3‰, whereas Tanganshan is somewhat less enriched at +29.3‰. For comparison, Tang and Liu (1999) reported shale δ34S values that average +53.5‰ and MnCO3 ores that average +49.3‰ for Minle. Li et al. (1999) reported two shale and two ore samples from Xiangtan that average +61.2‰ and +52.1‰ respectively, and two shales and four ore samples from Songtao that average +45.6‰ and +54.4‰, respectively. Thus not only are the Neoproterozoic ores peculiar in their high S values, they are also peculiar in the coincidence of the shale and MnCO3 values, which indicates a difference in oreforming conditions between the Neoproterozoic and the younger deposits. The Paleoproterozoic Kalahari ore samples were also analyzed for S isotopes, but the amount of S was too low for reliable measurement.
We did not measure C isotopes in our samples, but there are literature data from nearby deposits. Tang and Liu (1999), Li at al. (1999), and Yang et al. (1999) report 24 analyses of carbonate carbon, mostly from the Minle and Songtao deposits. The average δ13C value is −8.7‰ with a narrow standard deviation of 1.9. This value is similar to the average of −9.1‰ for the Kalahari deposits (Gutzmer, 1996) and −13.1‰ for Molango (Okita, 1987). Thus the Chinese deposits fall within the range for other MnCO3 ores for this parameter.
General Model for Mn Ore Genesis
In modern sedimentary basins, Mn shows a strong tendency to be enriched around the margins of areas with deeper, anoxic bottom waters. The Baltic and the Black Seas exhibit this behavior (see, e.g., Sternbeck and Sohlenius, 1997). The cause is the great insolubility of the Fe sulfide, pyrite, compared to the Mn sulfide, alabandite. Fe and Mn are transported into the basin of deposition as coatings of Fe2O3 or MnO2 on detrital particles and are released as soluble Fe2+ or Mn2+ to the pore waters of the sediment during diagenesis through bacterial reactions (see e.g., Potter et al., 2005, p. 138). The dissolved Fe is quickly incorporated into pyrite, but the Mn diffuses up into the overlying water mass. As a consequence, Fe is vanishingly low in the deep portions of anoxic basins, whereas dissolved Mn is much higher. Both substances form insoluble oxides and so are absent in solution in the shallow, oxygenated water mass. There is a peak in dissolved Mn just beneath the redox interface that reflects the redissolution of MnO2 particles that form in the shallow water and sink through the interface (Fig. 10).
From this cycling pattern of Mn in modern basins, a general model of Mn ore genesis has been developed in which Mn is solubilized from deep-water sediments in anoxic basins and reprecipitated around the margins of these basins at the point where the redox interface impinges on the seafloor. This model was first articulated by Force and Cannon (1988) from their observations of Mn distributions in modern sediments and facies analysis of several ancient Mn deposits. Subsequently the model has been developed in some detail based on stable isotopic studies of Phanerozoic deposits, particularly Molango in Mexico (Okita, 1987; Okita et al. 1988; Maynard et al., 1990; Okita 1992; Okita and Shanks, 1992). See also reviews by Force and Maynard (1991), who emphasized the ancient record and favored a dominant role for basin geometry, and by Calvert and Pedersen (1996), who emphasized the modern record and argued for a dominant role of surface-water productivity in controlling Mn distribution.
The process of Mn enrichment begins with the precipitation of Mn oxides within the water column at the interface between oxidizing and reducing conditions, usually a halocline. Most of the precipitated Mn simply redissolves as it passes downward through the water column, unless the seafloor is shallow enough to intercept the redox interface (Fig. 11). This phenomenon produces what might be called a “manganese compensation depth.” Below this depth, which occurs at about −200 m in the Black Sea, MnO2 particulates dissolve while settling through the water column and none reach the bottom. Above this depth, solid is stable and does not dissolve as it sinks. Consequently, the deep-water sediments are low in Mn, sediments close to the compensation depth have a strong enrichment in Mn oxide particles, and shallow-water sediments are low in Mn. Thus there is a critical depth for Mn enrichment that produces a “bathtub ring” around the margins of the basin. Where the clastic sedimentation rate is low, significant Mn accumulations can develop. Although these accumulations start as oxides, they are usually preserved as carbonates in the rock record. Reaction with organic matter in the sediment converts the primary Mn oxide to secondary Mn carbonate, which is depleted in 13C as a result of derivation of a portion of the carbon from decaying organic matter.
A key observation supporting this model is a strong correlation between Mn contents in the rocks and C isotopes (Okita and Shanks, 1992). The production of the MnCo3 mineralization requires the consumption of large amounts of organic matter and most likely occurred during early diagenesis, when bacterial processes are most effective. The process can be represented schematically by the reaction:
At the same time that the Mn is oxidizing the organic matter, it also attacks any Fe sulfide in the sediment (Aller and Rude, 1988; Schippers and Jørgensen, 2001). Pyrite is nearly ubiquitous in marine sediments because of the reaction between detrital Fe oxides and seawater sulfate. In Mn-rich sediments, however, this process is blocked because any precursor FeS that forms is quickly destroyed:(2) predicts that Mn ore deposits should be very low in S, as is observed in most occurrences other than those in the Neoproterozoic. Also the minor amount of pyrite that does form should be relatively heavy isotopically. As mentioned above, this prediction of heavy S is based on the requirement that any pyrite that forms be relatively late, forming after all of the Mn oxide has been converted to MnCO3. Therefore the degree of contact with the overlying seawater reservoir of sulfate S will be limited, sulfate reduction will go to completion, and the small amount of sulfide that does form will be isotopically close to its parent sulfate, in contrast to normal pyrite in black shale, which is highly depleted in 34S. Subsequent work has shown that this model has broad applicability to Mn ore deposits. See for example Nyame (1998) on the Nsuta deposit of Ghana, and Tsikos (1999) for the Hotazel deposit in the Kalahari Mn field of South Africa.
Ore Genesis in the Neoproterozoic—The Snowball Earth Model
The Neoproterozoic deposits appear chemically to be similar to their older and younger counterparts, except for different S isotope and REE patterns. The similarity in carbon isotopes and trace element chemistries suggests that the MnCO3 formed during early diagenesis from a Mn oxide precursor, just as in other deposits. How was this Mn oxide deposited in such large amounts and what accounts for the super heavy S in the associated pyrite?
An appealing hypothesis has been put forward by Gorjan et al. (2000; see also Gorjan et al., 2003) in which they related isotopic behavior and the abundance of Fe and Mn mineralization to turnover after melting of glacial ice on a “snowball” Earth (Fig. 12). The concept of a frozen Earth in the Neoproterozoic (Kirschvink, 1992) explains many of the biological, sedimentological, and geochemical anomalies seen at this time and accordingly has received much attention. In the Gorjan et al. (2000, 2003) version of this model, the ocean became totally anoxic under its ice cover, which led to a buildup of dissolved Mn and Fe in the bottom water. Because the flux to the oceans of detrital Fe from the continents and hydrothermal Fe from ridge crests is much greater than the flux of dissolved SO4 2- and because the seafloor was everywhere anoxic, pyrite formation in sediments or at the ridges removed most of the sulfur, but left plenty of Fe to accumulate in the water column as soluble Fe2+. This situation is the reverse of the modern oceans, where oxic bottom waters keep most of the detrital Fe insoluble as the oxide and quickly convert any soluble Fe2+ in vent fluids into insoluble Fe3+. As a consequence, dissolved S is in great excess over dissolved Fe, and the SO4 2- is high.
The precipitation of most of the seawater reservoir of sulfur as sulfide resulted in the sequestration of large amounts of 32S in sediments and the accumulation of residual 34S-enriched S in the water column. On melting of the ice, there is overturn of the oceans and this Fe- and Mn-laden deep water wells up onto shallow platforms and deposits the Rapitan-type iron formations characteristic of the Neoproterozoic (Maynard, 1991) and large deposits of Mn oxide. Because this upwelling deep water contains sulfate strongly enriched in 34S, the shallow-water diagenetic pyrite produced from it is similarly very heavy.
Although in broad terms this model provides a satisfactory explanation for a wide array of observations, there are some significant discrepancies:
The Fe and Mn deposits are interbedded with the glacial deposits rather than succeeding them.
The ores formed in narrow rifts rather than on open shelves.
Sulfur isotopic compositions are typically heavy throughout the Proterozoic and excursions appear to be unrelated to glacial episodes.
The REE evidence from the associated iron formations is incompatible with a totally anoxic deep ocean.
An Ice-covered Rift Model for Ore Genesis in the Neoproterozoic
These aspects of the Mn-Fe mineralization in the Neoproterozoic suggest that the snowball Earth model needs to be refined. The hypothesis that best explains these additional observations is a “partial snowball Earth,” with ice-covered continents and marginal seas, but with an ice-free open ocean at low latitudes. See Eyles and Januszczak (2004a, 2004b), Poulsen et al. (2002), and Poulsen (2003) for useful discussions of full versus partial ice cover from a field and from a modeling basis. Three principal lines of evidence support a partial, or “soft” snowball model for mineralization: stratigraphic sequence, stable isotopic values, and REE compositions.
Stratigraphic Setting of Mineralization
The interglacial position of the ores suggests that the snowball Earth model of Fe-Mn deposition needs to be modified to allow the main episode of mineralization to occur during periods of partial melting of the ice cover accompanied by oxidation of surface waters. Multiple glacial advances and retreats are hard to reconcile with a totally frozen Earth. Such fluctuations are, however, compatible with a model of individual ice-covered marine basins that could have been time transgressive, as proposed by Eyles and Januszczak (2004a, 2004b).
The dominance of glacial-marine rift environments is also more consistent with partial ice cover. The Neoproterozoic Fe and Mn deposits are uniquely associated with narrow rifts on continental crust (Maynard, 1991); none are found on stable shelves of continental margins. The “hard snowball” explanation for the Fe and Mn mineralization would predict deposition as widespread sheets on continental shelves (see Gorjan et al., 2003, p. 95), much like the famous Pennsylvanian black shales of the U.S. midcontinent described by Heckel (1991), which are related to maximum flooding associated with deglaciation.
Sulfur Isotopic Values Throughout the Neoproterozoic
The recent compilations of sulfide S isotopic data by Strauss (1997) and by Canfield and Raiswell (1999) show that heavy sulfur has been characteristic of both sulfides and sulfates during much of the Proterozoic and is not correlated to the glacial episodes. Another important aspect of the Proterozoic S record is that the spread between sulfate and sulfide values is much less than in Phanerozoic rocks, with several sulfide analyses lying higher than contemporaneous seawater sulfate. Strauss (1997) suggested that these patterns could result from a low-sulfate ocean in which the amount of pyrite formed at the sediment-water interface was limited by the amount of sulfate in the overlying water rather than by the amount of organic carbon, which is the case in modern marine sediments (e.g., Canfield, 2001). A low-sulfate ocean would have been prone to periodic drawdowns of sulfate concentration during times of higher Fe flux to the ocean, which would have resulted in increased pyrite formation. The isotopic composition of the residual sulfate S in seawater would then have spiked to very high values.
Determining a worldwide curve for Neoproterozoic S isotope values comparable to the familiar curves for the Phanerozoic has been difficult because of the scarcity of primary sulfate minerals from this time period and the high variability in sulfide values for rocks of a given age. We have made an approximate sulfide S curve by plotting the medians of the values reported by Strauss (1997) for each time interval. As shown in Figure 13, the median values show large fluctuations, but are invariably heavy compared to Phanerozoic values. There do seem to be periods within the Neoproterozoic that experienced spikes to extremely heavy sulfide S, but these do not match the main glacial episodes. Instead there seems to be a sharp fall in δ34S following each glaciation, the converse of the prediction from the “hard snowball” model. Perhaps these swings indicate periods of increased deep-water circulation that produced well-oxygenated oceans with SO4 2- concentrations.
A finer-scale sulfate S curve is available from the work of Hurtgen et al. (2002), who measured trace sulfate held in carbonate minerals from the Neoproterozoic of Namibia. Their sampling starts above the Sturtian glacial strata and thus does not cover the interval corresponding to the Chinese Mn deposits, but their results do provide important constraints on the mechanisms for Neoproterozoic S change. They identified four major excursions to heavy S: one is slightly above the Sturtian glacials and reaches +40‰; the second is 150 m higher in the section and has the greatest departure, to +50‰; the third is midway between the Sturtian and Marinoan glacial episodes and reaches +35‰; and the final excursion is 100 m above the Marinoan equivalents and reaches +40‰. Thus there are several episodes of development of heavy S and only the first one can be convincingly related to glaciation. Their data also shows that the sulfate δ34S values are comparable to the sulfide values, indicating that sulfate reduction within sediments must have gone essentially to completion during this time interval. Hurtgen et al. (2002) and Canfield (2004) interpreted these patterns to indicate generally low SO4 2- concentrations in seawater at this time. Canfield (2004) has suggested values as low as 200–300 µM, which he attributed to a much greater flux of sedimentary S back to the mantle via subduction of pyrite-rich deep-sea sediments during the Proterozoic than in the modern oceans.
We conclude that the occurrence of extremely heavy δ34S values of pyrite in the Chinese Mn deposits is not related directly to glaciation but to generally low concentrations of dissolved SO4 2- in the Neoproterozoic world ocean that made seawater subject to rapid and severe swings in its S content. Excess iron, which must have been present judging from the abundance of Rapitan-type iron formations, would have driven SO4 2- concentrations to very low values through formation of more pyrite (Canfield and Raiswell, 1999), and this could account for the uniformly extremely high δ34S values found. The combination of high Fe flux and a low SO4 2- ocean is what produced the S isotopic signature of the Chinese Mn deposits. We further conclude from our survey of literature data that these very heavy values occurred repeatedly at many times and in many places in the Neoproterozoic and are hence more likely to have resulted from world ocean effects rather than from peculiar chemistry in isolated basins.
REE Evidence for an Oxidizing World Ocean
The nature and source of this Fe flux can be reconstructed from the behavior of REE in the Mn deposits and in their iron formation cousins. Iron formations have distinctive REE patterns for each of the three main periods of iron mineralization (Fig. 14). Both the Archean Algoma-type and Paleoproterozoic Superior-type iron formations have pronounced positive Eu anomalies on NASC normalized plots. Notice also in Table 3 that all of the NIST iron-ore samples, which are from the Lake Superior region, have positive anomalies. This anomaly is conspicuously absent in the Neoproterozoic Rapitan-type deposits (Derry and Jacobsen, 1990; Klein and Beukes, 1993; and Bau and Möller, 1993; Klein and Ladeira, 2004).
Positive Eu anomalies are associated with mid-ocean ridge vent fluids (see, e.g., Cocherie et al., 1994, on Red Sea sediments). Destruction of calcic plagioclase in the oceanic crust leads to a release of excess Eu to hydrothermal solutions. Because of the high temperatures required for the removal of Eu from plagioclase, Eu release is confined to axial vents. The results of Michard and Albarède (1986) suggest that temperatures greater than 350 °C are required. Furthermore, at lower temperatures, there seems to be little release of any of the REE seawater. For example, Wheat et al. (2002) studied low-temperature hydrothermal springs from the Juan de Fuca ridge and found them to be net sinks instead of sources of REE to seawater. It also seems likely that most Mn release is from the axial vents: Murton et al. (1999) were able to account for all of the Mn release from a 50-km segment of the Mid-Atlantic Ridge by flux from the Broken Spur vent field.
Normally, these vent-sourced REE are immediately scavenged by Fe oxides precipitating around the vents (Mitra et al., 1994). If, however, the ridge-crest hydrothermal systems vent into oxygen-free bottom water, this precipitation of Fe oxides and scavenging of REE will not occur, and both the Fe and the vent-signature REE can be concentrated in bottom waters and carried into shallower waters to be precipitated far from their source. The evolution of REE patterns with age suggests that the Archean and Paleoproterozoic iron formations had Fe dominantly sourced from ridge-crest vents discharging into anoxic seawater, whereas the Neoproterozoic deposits received their Fe from seawater with a relatively minor hydrothermal component. The absence of a positive Eu anomaly in the Neoproterozoic deposits indicates that they were deposited during periods of general oxidation of oceanic bottom water. Dobrzinski et al. (2004) came to the same conclusion using a variety of paleoredox indicators.
Neodymium isotopes also suggest a waning hydrothermal influence on Fe deposits through time. Jacobsen and Pimentiel-Klose (1988) reported that Archean and Paleoproterozoic iron formations have ϵd values similar to the mantle, whereas the Neoproterozoic Urucum iron formations are similar to modern seawater. The average values are +2.7 for Algoma-type IF, +1.0 for Superior-type, and −2.9 for a single determination from Urucum (reported in Derry and Jacobsen, 1990).
Eu anomalies in Mn deposits do not present such a clear picture (Maynard, 2004). Note in Table 3 that virtually all Mn deposits, regardless of age, have positive Eu anomalies. This may indicate that a significant hydrothermal contribution is present in Mn ores throughout geologic time, and furthermore that the bottom water into which the volcanic solutions exhaled was anoxic, as in the Force and Cannon (1988) model. The Chinese deposits are mixed, with the three Tanganshan ore samples each having a positive anomaly whereas all Xiangtan ore samples have a negative anomaly. Neoproterozoic Mn deposits in the Urucum district of Brazil have uniformly negative anomalies (Graf et al., 1994), similar to seawater. The Xiangtan and Urucum deposits may have a much lower hydrothermal component than the other Mn ores or perhaps the bottom water in their basins was sufficiently oxic to permit formation of Fe oxides close to the vents, scavenging all of the vent-derived REE.
The behavior of Ce may provide some clues to this situation. Ce shows a much stronger association with the Mn component of Fe-Mn accumulations than to the Fe component (DeCarlo, 1991). Fleet et al. (1983) has suggested that the relative size of the negative Ce anomaly on NASC-normalized plots has a linear relationship to the ratio of hydrothermal- to hydrogenous-sourced material in the Mn accumulation. In the modern oceans, the hydrothermal end member has a strongly negative anomaly with a Ce/Ce* value of ∼0.46 (Hein et al., 1996, their table 8 corrected for geometric calculation of Ce*). Hydrogenous Mn accumulations, sourced only from seawater, have a strongly positive Ce anomaly with a value of ∼1.6 (data of Usui and Someya, 1997). The rate of deposition accounts for much of this difference (Maynard, 2004). The hydrothermal deposits form relatively quickly and preserve the original REE signature of the water, whereas the hydrogenous Mn nodules form extremely slowly and, in the modern oxygen-rich deep sea, catalyze the oxidation of Ce3+ to Ce4+, producing significant Ce enrichments and high Ce/Ce* values. Almost all Fe and Mn ores in Table 3 show negative Ce anomalies, indicating an appreciable hydrothermal contribution into anoxic bottom waters. The Chinese deposits depart from this trend, being uniformly positive like the modern Mn nodules, suggesting at least mildly oxidizing conditions for the Neoproterozoic and slower deposition rates than in other Mn deposits. Slower deposition is also consistent with the higher ΣREE seen in the Chinese deposits (Table 3).
The combination of the Eu and Ce data indicate that Archean and Paleoproterozoic Fe and Mn deposits are dominantly hydrothermal in the source of metals and REE and that the ocean basins worldwide were dominantly anoxic at the time of their formation. The Neoproterozoic Fe deposits, and possibly the Mn deposits, received a minor hydrothermal contribution and were deposited during a time of global oxygenation of the oceans. In the Phanerozoic, iron formations are confined to the immediate vicinity of hydrothermal vents, whereas Mn is far-traveled. Each Mn deposit in Table 3 seems to have had a variable hydrothermal component, but each was associated with a large but isolated basin with anoxic bottom water (e.g., Maynard et al., 1990; Force and Maynard, 1991).
If seafloor hydrothermal processes were supplying only limited amounts of Fe and Mn to the Neoproterozoic ore deposits, what was the source of the high metal fluxes? We suggest that low-latitude glaciation scraped off tropical soils highly enriched in Fe and Mn in the form of oxides rather than silicates and thus highly susceptible to diagenetic remobilization. This lateritic material was then dumped into small ocean basins that had limited communication with the open ocean and so could become anoxic at depth without receiving the mid-ocean ridge signature of REE. Mn was then exported to shallow water to precipitate at the oxic/anoxic interface, while some Fe precipitated in deep SO4 2- in the basin. This process left residual sulfur strongly enriched in 34S to be incorporated in the shallow-water deposits. Thus the super-heavy values we find are not the direct result of glaciation but are an indirect result through the impact of glaciation on Fe behavior (Fig. 15). Support for a lateritic source of the Fe and Mn comes from the very low Na2O content of the host shales which averages 0.14% at Tanganshan and 0.07% at Xiangtan, compared with typical shales, which have ∼1% Na2O (Li, 2000, table VI-4). Dobrzinski et al. (2004) used the CIA index of Nesbitt and Young (1982) to characterize the weathering state of the source area for the Neoproterozoic deposits of south China. They calculated an average CIA of 62 for the lower diamictite, 71 for the interglacial unit, and 65 for the upper glacial unit. For comparison, modern glacial marine sediments of the Scotia Sea average only 55 (data of Diekmann, et al. 2000), whereas modern soils average 72 (Maynard, 1992). Thus the Chinese Neoproterozoic glacial section shows higher degrees of weathering than the modern, and the Datangpo Formation was sourced from deeply weathered material.
This model of Neoproterozoic Fe-Mn mineralization raises several questions that cannot be answered with the data presently available:
Why did extensive Fe and Mn mineralization not occur during the second glacial episode of the Neoproterozoic? The available literature indicates that all the sizable occurrences are Sturtian. Are there mis-correlated Marinoan-age deposits or are they truly absent? Could the continents have lacked an extensive lateritic weathering mantle at the time of the Marinoan glaciations? Perhaps the Marinoan glacial episode was truly global whereas the Sturtian was only partial.
What causes the variation in Eu anomaly in Mn ores, but not in the Fe ores? Does it reflect more localized sources for Mn than for Fe? Does it reflect preferential sorption of Eu to Mn oxides under certain conditions?
What causes the other S isotope excursions in the Neoproterozoic that are not associated with Fe-Mn deposits? Are these times of higher Corg input? Could they be times of evaporite deposition with drawdown of sulfate levels? Could they be times of enhanced Fe flux from mid-ocean ridges? Could they reflect times of greater global anoxia in which more of this mid-ocean ridge flux goes to pyrite rather than to Fe oxide?
Resolving these questions will require detailed geochemical analyses of a range of Neoproterozoic deposits. In particular, the analyses of small amounts of sulfate δ34S contained in carbonates (e.g., Kampschulte and Strauss, 2004) need to be extended to older rocks and to other sections, and more REE data are needed.
The Neoproterozoic Mn deposits of China were deposited in a restricted, linear basin during partial retreat of the Sturtian glaciers. Their exceptionally heavy δ34S values can be explained by an excess of Fe flux over S flux to these restricted basins. Eu anomalies in associated Fe ores indicate that this Fe flux was dominated by diagenetic rather than hydrothermal sources, whereas Ce anomalies in Mn ores suggest relatively oxidizing conditions in the bottom water at the site of Mn deposition. A possible source of enhanced diagenetic flux of Fe and Mn is rapid deposition of lateritic soil residues scraped off by low-latitude glaciers into sub-oxic or anoxic bottom waters of a rift basin. The easily remobilized Fe and Mn oxides in this residue would have accumulated in low-oxygen bottom water and then precipitated under more oxidizing conditions around the edges of the basin during sea-level rise in interglacial periods.
Figures & Tables
Evolution of Early Earth's Atmosphere, Hydrosphere, and Biosphere - Constraints from Ore Deposits
- anaerobic environment
- ancient ice ages
- banded iron formations
- black shale
- chemical composition
- chemically precipitated rocks
- clastic rocks
- Far East
- glacial geology
- Hunan China
- interglacial environment
- iron formations
- isotope ratios
- manganese ores
- metal ores
- mineral deposits, genesis
- organic carbon
- rare earths
- sedimentary rocks
- snowball Earth
- stable isotopes
- sulfate ion
- upper Precambrian
- whole rock
- X-ray fluorescence spectra
- southern China
- Taojiang Deposit
- Minle Deposit
- Xiangtan Deposit
- Tanganshan Deposit