Introduction

Together with Sr isotopes, the Pb isotopes of OIB showed, for the first time, that the mantle is heterogeneous1, most likely due to the presence of enriched mantle 1 (EM1), enriched mantle 2 (EM2) and HIMU end-components from recycled crust and sediment2,3. High μ OIB from Saint Helena, Mangaia and Tubuai islands have unusually high 206Pb/204Pb (>20.0) and low 87Sr/86Sr (≤0.7028), and the bulk of OIB and many MORB also have high 206Pb/204Pb although higher 87Sr/86Sr (>0.7028). That is, almost all oceanic basalts have high 206Pb/204Pb that plot to the right of Geochron, a line in the 206Pb/204Pb versus 207Pb/204Pb diagram containing the current Pb isotopes of terrestrial materials, assuming the Earth remained a closed system (Fig. 1A). Both 206Pb and 207Pb increase due to the respective decay of radioactive 238U and 235U and, hence, the mantle source of oceanic basalts must have a long time-integrated (~b.y.) U/Pb > BSE refs2,3,4. Such a high U/Pb is unexpected because U is more incompatible than Pb and, thus, the ratio should be low at least in the depleted MORB mantle (DMM) source that has experienced repeated melt extraction. This has been termed the main or 1st Pb paradox, and its proposed solutions include transfer of Pb into the core4,5, preferential retention of Pb relative to U in the continent6,7 or residual mantle sulfide8, hydrothermal transfer of Pb from mantle to continent9,10, U recycling into the mantle since Early Proterozoic2,11, and crustal contamination of an early-formed (ca. 4.5 Ga) mantle reservoir12. Notably, a majority of these solutions call for a decrease in Pb to increase U/Pb refs2,3. Moreover, the distinction between the sources of HIMU end-component, which has characteristically low 87Sr/86Sr, and HIMU-like characteristics of many oceanic basalts, which have variable and higher 87Sr/86Sr, has become blurred. A majority of the proposed solutions do not consider that the HIMU end-component is distinct because it must also have a long time-integrated, low Rb/Sr, as the decay of radioactive 87Rb increases 87Sr.

Figure 1
figure 1

206Pb/204Pb versus A. 207Pb/204Pb and B. 208Pb/204Pb for Archaean limestones. Bulk stromatolitic limestones (solid symbols) are connected by tie lines to their respective carbonate fractions (open symbols). Analytical errors are smaller than the symbols used. Shown for reference are the carbonate fractions of other Archaean limestones (see text for sources of data), oceanic basalts (gray field), FOZO (dash area inside gray field), and HIMU end-component (dark gray field). The 2.82 Ga isochron in A. is the regression line for the extremely radiogenic (206Pb/204Pb = 30.2 to 104.4; 207Pb/204Pb = 19.4 to 34.1; 208Pb/204Pb = 42.514 to 74.402) Mushandike stromatolites24 (not shown).

Full size image

Measured κ (2.5 average) for most MORB are lower than those deduced from their 208Pb/204Pb and 206Pb/204Pb, which are fairly homogeneous and only slightly lower than bulk silicate Earth (BSE; κ ca. 4.0) refs13,14. The increase in 208Pb results from the decay of radioactive 232Th. The measured low κ of MORB indicates that the Th/U decrease from the original BSE κ in DMM may have occurred as recently as ca. 600 Ma ref.13. However, the Th-enriched continental crust derived from DMM is much older than this and, thus, the Th/U evolution of the mantle is a dilemma2,13,14. This has been termed the 2nd Pb paradox, and its proposed solutions include a mantle BSE κ that decreased concurrently with continental extraction15,16, an upper mantle BSE κ that abruptly changed to 2.5 at 600 Ma13 and mantle recycling of U due to the Earth’s oxidizing condition since Early Proterozoic2,11,14,17.

Quite recently, we hypothesized that some Archaean marine carbonates provide the radiogenic 206Pb/204Pb, unradiogenic 87Sr/86Sr, low K2O, and high CaO of HIMU, as well as the overall positive correlation between CaO and 206Pb/204Pb in OIB ref.18. Our hypothesis is consistent with the Archaean-formed Δ33S anomaly in HIMU OIB ref.19 and the similar trace element contents of carbonatitic melt inclusions in diamonds and HIMU lavas20. We also proposed that the unusually radiogenic Pb isotopes of OIB (Fig. 1A) and 1st and 2nd Pb paradoxes are parts of a system of equations, and the higher concentration of U relative to Pb and Th in marine carbonates offers a self-consistent solution to these equations21. Such a holistic interpretation of Pb isotopes, however, assumes some Archaean marine carbonates formed in equilibrium with seawater containing U, but little to no Pb and Th, as well as high Sr but little to no Rb ref.18. This assumption is inconsistent with the anoxic Archaean ocean that should have prevented U mobilization in seawater and its eventual subduction into the mantle2,11,14,17.

To verify if some Archaean marine carbonates have high U/Pb and U/Th, we analyzed the Pb and Sr isotopes of the carbonate fractions of the few stromatolitic limestones that have the low 87Sr/86Sr of coeval seawater22,23. Notably, these carbonates preserve the composition of Archaean oceans, have long time-integrated low Rb/Sr and are generally unaltered, as alteration would have raised their 87Sr/86Sr refs22,23. Accordingly, the Pb isotopes of these unaltered samples should also be able to constrain their long time-integrated U, Th and Pb concentrations and ratios. We adopted this approach because the current U, Th, Pb, Rb, and Sr concentrations of many Archaean limestone outcrops have been modified by contamination and alteration22,23,24,25,26. We also analyzed their bulk Pb and Sr isotopes for comparison. Four samples (184, 185, 1977/7, and 1977/43) are from the 2.8 Ga Mosher Carbonate Formation in the Canadian Shield27 and one sample (187) is from the 3.0 to 2.6 Ga Bulawayan Supergroup in the Zimbabwe Craton28. The location and full description of the samples can be found in ref.23.

Results

The bulk limestone Pb isotopes are highly variable, but their carbonate fractions have systematically higher 206Pb/204Pb and 207Pb/204Pb (Table 1; Figs 1 and 2). On the other hand, the carbonate fractions of samples 184, 185 and 1977/7 have systematically lower 87Sr/86Sr than their respective bulk values whereas those of samples 187 and 1977/43 are similar. The 87Sr/86Sr of the carbonate fractions of samples 187, 1977/7 and 1977/43 and their previously reported values23 are the same within error despite the difference in the leaching procedure used, and are also similar to the recent estimate of Archaean seawater 87Sr/86Sr ref.29.

Table 1 Lead and Sr isotopic ratios of Archaean limestones with the low 87Sr/86Sr of coeval seawater.
Full size table
Figure 2
figure 2

206Pb/204Pb versus 87Sr/86Sr for Archaean limestones. Symbols as in Fig. 1. Shown for reference are the carbonate fractions of the few Mushandike limestones that have 87Sr/86Sr data (gray arrows point to the approximate location of other samples), the field for oceanic basalts and locations of proposed mantle end-components.

Full size image

The 206Pb/204Pb and 207Pb/204Pb of bulk and carbonate fractions of our samples plot with or at the extension of the wide 206Pb/204P-207Pb/204Pb array of the carbonate fractions of Manjeri limestones24 that also occur in the Zimbabwe Craton and of Nsuze limestones26 and Transvaal dolomites30 in the Kaapvaal Craton (Fig. 1A). Notably, oceanic basalts plot within this array. The array, in turn, plots below the arrays of the carbonate fractions of Cheshire limestones25 as well as of surface (plotted) and very radiogenic drill core (not shown) samples from Mushandike stromatolites24, all in the Zimbabwe Craton. In detail, the 206Pb/204Pb and 207Pb/204Pb of three Mosher limestones (184, 185, 1977/43), together with some of the Transvaal dolomites, are more radiogenic than OIB. Finally, the 208Pb/204Pb for given 206Pb/204Pb of bulk limestones, their carbonate fractions and those of many Archaean limestones are highly variable, but for many samples they are lower than those of oceanic basalts (Fig. 1B). Significantly, some Cretaceous serpentinites and 1.9 Ga altered oceanic crust have unusually low 208Pb/204Pb, clearly indicating the extremely low Th content and, hence, Th/U of oxygenated seawater18,31.

The systematically high 206Pb/204Pb plus 207Pb/204Pb and low 208Pb/204Pb of the carbonate fractions we analyzed clearly indicate that some Archaean limestones have long time-integrated high μ and low κ refs18,21. A possibility is that such limestones precipitated locally in Archaean ‘oxygen oases’ in shallow-platforms27,32. Indeed, the very radiogenic Pb isotopes of the aforementioned Archaean Mushandike limestones were most probably generated in a shallow restricted basin with variable, but limited communication with the open ocean24. Alternatively, they may have been formed during localized concentrations or ‘whiffs’ of oxygen in the Archaean33, as proposed for the Nsuze metasediments34. Whichever the case, marine carbonates that precipitated in equilibrium with oxygenated seawater theoretically can have extremely high μ and low κ ref.35. Accordingly, the radiogenic Pb isotopes of some Archaean carbonates and interlayered metasediments are most probably due to in-growth of uranogenic 206Pb and 207Pb in their carbonate fraction, rather than from contamination by a high μ Archaean crust24,25,26,36.

Discussion

Although there is consensus that the HIMU, EM1 and EM2 end-components are recycled surface material, there is little agreement on the origin of individual end-components. A currently popular idea is that HIMU is from altered MORB whereas EM1 and EM2 are from pelagic and terrigenous sediments, respectively2,3. Irrespective of sources, however, the end-components almost always mix with a ‘common’ or focus zone (FOZO) component2,3,37. The marine carbonate recycling hypothesis not only offers an alternative origin for the HIMU signature, but also explanations for some common threads in the geochemistry of oceanic basalts, particularly their unusually high 206Pb/204Pb and 207Pb/204Pb (Fig. 1A) and the almost ubiquitous link between OIB end-components and the HIMU-like FOZO2,3,18,37. Our new analyses for a few select Archaean limestones lend some credence to the hypothesis.

Sample 1977/43 and its carbonate fraction as well as those of samples 184 and 185 plot at the extension of HIMU OIB in the 206Pb/204Pb and 87Sr/86Sr diagram (Fig. 2). Although such limestones appear scarce, it is noteworthy that the measured Pb and Sr isotopes of many Archaean limestone outcrops may be secondary, as these are mainly effects of later metamorphism and/or alteration22,23,24,25,26. That our Archaean carbonate fractions still show their original μ, κ and Rb/Sr features is most likely due to a combination of favorable preservation and the sluggishness of element diffusion in solid rocks38,39. Significantly, if a subducted Archaean oceanic slab containing carbonates and/or carbonated metasediments would partially melt in the deep mantle, carbonate-rich melt would most likely form first since calcareous metasediment has the lowest melting temperature of all lithologies in the subducting lithosphere40,41. Moreover, high modes of melting would occur once melting initiates, as limestone is mainly comprised of a single calcite phase42. The carbonate-rich melt could then infiltrate and/or metasomatize the lithospheric mantle portion of the slab, as carbonate has higher diffusion rate than basaltic melt43,44. Subsequent melting of carbonate-metasomatized ancient upper mantle could produce OIB melts18,21,41.

To illustrate the origin of the 206Pb/204Pb and 207Pb/204Pb of HIMU OIB, we first calculated the evolution of the carbonate fraction of sample 185 (Fig. 3A). Sample 185 formed 2.8 Ga in the prevailing environment with a μ of 7.99 refs7 but its carbonate fraction evolved with a μ of 20.63 (dash line). We assumed the sample was part of a subducted slab that was isolated for quite some time (ca. ~b.y.) deep in the mantle to allow uranogenic 206Pb and 207Pb to grow. Then the carbonate partially melted to form a carbonatite melt that metasomatized the mantle portion of the subducted slab, which partially melted quite recently to produce plume magmas. Using these results and assumptions, we then constructed a model using a hypothetical limestone that evolved like sample 185 and relevant experimental data41. Metasomatism (solid line) of a subducted, evolving upper mantle7 (99%) by the partial melt from the limestone (1%) produced HIMU OIB containing a mere ≤20% of such melt. Notably, the metasomatic melt was highly enriched in 206Pb and 207Pb as it came from a limestone that only had U but no Pb. Moreover, the subducted slab was very low in Pb to begin with as its Pb was sequestered during subduction2,18 and/or later by mantle sulfides8, making the slab susceptible to contamination or spiking by the uranogenic Pb refs18,21. On the other hand, the low 87Sr/86Sr of HIMU OIB is simply that of the Archaean limestone, which had the 87Sr/86Sr of coeval seawater that, in turn, was being buffered by the upper mantle22,23 (Fig. 2).

Figure 3
figure 3

206Pb/204Pb versus A. 207Pb/204Pb and B. 208Pb/204Pb diagrams showing the evolution of the carbonate fraction of sample 185 from the 2.8 Ga Bulk silicate Earth7 to present (dash lines), and our model for the origin of HIMU OIB (solid lines). The model was constructed by metasomatizing (mixing) 99% of a slightly younger, 2.4 Ga evolving upper mantle (206Pb/204Pb = 13.728, 207Pb/204Pb = 14.614, 208Pb/204Pb = 33.210, 0.051 ppm Pb) ref.7 with 1% of a hypothetical Archaean limestone (206Pb/204Pb = 24.404, 207Pb/204Pb = 16.200, 208Pb/204Pb = 43.00; open square – M) containing the Pb of the most enriched, 8% partial melt from carbonated pelite (67.2 ppm; see ref.41 for details of experimental data, and modeling). The latter Pb concentration was used because of the dearth of experimental data for partial melting of limestone at high pressures. Each upper mantle and total silicate Earth symbol represents 0.4 Ga interval whereas each tick mark along the model mixing line represents 0.2 increment of melt contribution. See text for additional discussion.

Full size image

Our model illustrates a similar, idealized origin for the 208Pb/204Pb of HIMU OIB (Fig. 3B), but this is an over-simplification. Sample 185 formed in the prevailing environment with a κ value of 3.99 ref.7 but evolved with a lower value of 2.49 (dash line), similar to many Archaean limestones (Fig. 1B) and/or serpentinite and altered crust31 that have extremely low κ values. However, if our hypothetical Archaean limestone indeed fully-equilibrated with oxygenated seawater, then the carbonatitic melt contribution to the 208Pb/204Pb of HIMU OIB is negligible because it had none to very little thorogenic 208Pb and, compared to Pb, Th was sequestered much less from the slab during subduction. Thus, the 206Pb/204Pb-208Pb/204Pb evolution of HIMU OIB (solid line) is primarily an extension of the upper mantle7. This notion is consistent with geochemical data indicating a uniform Th/U ratio for the ancient upper mantle7 and modern DMM refs21,45,46.

Our model also implies that mixing between an even smaller amount of Archaean (or perhaps younger) limestone and ancient upper mantle can produce the variable and radiogenic Pb and Sr isotopic ratios of the FOZO component18,21. On the other hand, the HIMU-like 206Pb/204Pb and radiogenic 87Sr/86Sr of the bulk of OIB can be produced through mixing between ancient mantle and carbonated crust plus sediment (e.g., containing ankerite23). In this case, mixing between ancient mantle (FOZO) and carbonated, variably altered MORB plus mafic (EM1) ref.47 to silicic (EM2) ref.48 metasediment are most likely involved (see also ref.41). Finally, mixing between carbonatitic melts from the present-day down-going slab that intersects the solidus for carbonated MORB at ca. 300 to 700 km ref.49 and DMM can produce the radiogenic Pb and Sr isotopes of modern enriched-MORB. Thus, recycling of marine carbonates can also solve the 1st Pb paradox21.

We additionally propose that the Th deficiency in many ancient carbonates produced the limited variation of the κ values of OIB that hover around the BSE κ, as these are primarily extensions of the evolving upper mantle (Fig. 3B). The homogeneous DMM κ value deduced from MORB 206Pb/204Pb and 208Pb/204Pb is also an upper mantle signature2,3,14,16. However, the fairly recently decoupled low Th/U of MORB (i.e., 2nd paradox)13 comes from the aforementioned carbonatitic melts from the down-going slab18,49 that are mainly from Proterozoic to Phanerozoic limestone, as melts from carbonated sediment has high Th/U ref.41. The occurrence of carbonatitic melts containing U in DMM is consistent with the isotopically distinct, high 238U/235U of MORB ref.17. The ‘exotic’ 238U/235U is formed in the oceanic crust at the bottom of the modern oxic ocean and subducted into the mantle, but must have been recycled back relatively quickly to the surface17.

In summary, our results indicate the existence of Archaean carbonates with the appropriate Pb and Sr isotopes to generate the HIMU source, per the marine carbonate recycling hypothesis. Through qualitative modeling, we illustrate via mixing between recycled upper mantle and partial melts from 1) select ancient marine carbonates the compositional signature of HIMU OIB, and 2) ancient carbonated crust and sediment the solutions to both 1st and 2nd Pb paradoxes. Notably, a recent study also concludes that the high U/Pb and U/Th content of ancient seawater as the source of the Pb isotopic signature of the HIMU end-component31. Instead of Archaean carbonatitic melt, however, the study proposes metasomatism of the source with U-enriched, supercritical liquid derived from Proterozoic (1.9 Ga) seawater. Incidentally, Proterozoic seawater would also impart a higher 87Sr/86Sr (>0.704)22,23 to the HIMU source. The study also locates the HIMU source domains roughly in the mid-mantle, where they can be fortuitously entrained by upwelling mantle plumes31. We acknowledge that the hypothesis is inconsistent with some existing data and prevailing interpretations. Foremost among these are that carbonates should break and dissolve completely during subduction50,51, there are very few Archaean carbonate outcrops that posses the putative HIMU composition, and carbonate melts41 and carbonatites52,53 are generally depleted in Zr, Hf and Ti but these are not possessed by HIMU OIB ref.2. However, there are also data indicating the occurrence of carbonate-sourced diamonds deep in the mantle54,55 and that carbonates can be preserved to great mantle depths56. Moreover, HIMU OIB is only a small fraction of OIB ref.3. Caution must also be taken when equating the current outcrops as true representations of the actual amount of subducted Archaean carbonates. Finally, our model indicates that the HIMU plume source contains only a minute amount of carbonatitic melt, and higher degrees of partial melting of a silicate-dominated source should increase the Zr, Hf and Ti contents of the resultant silicate melt41,57. Thus, more data are needed to test our hypothesis.

Methods

The limestone samples were broken into cm-sized fragments, washed with distilled water, dried, and pulverized in an agate mortar and pestle. About 50 mg of the limestone powders were completely dissolved using a mixed, ultra-pure HF:HNO3 solution in Teflon beakers and then analyzed for bulk limestone Pb and Sr isotopic composition following the established sample purification and thermal ionization mass spectrometry procedure at the Scripps Institution of Oceanography58. To analyze the Pb and Sr isotopic composition of carbonate fractions, ca. 100 mg of powders were leached with 3 ml 1 M ultra-pure acetic acid in Teflon beakers, ultrasonicated for 15 minutes and left to react at room temperature for ca. 24 h. Then the mixtures were centrifuged and supernates were carefully separated and dried down. Afterwards, ca. 0.5 ml ultra-pure, concentrated HNO3 was added to the supernates to oxidize any organics. The samples were dried down again prior to Pb and Sr purification and isotopic analysis following the same above procedure. Note that the separation procedure for carbonate fraction used in this study differs from that employed in refs22,23 in the use of mild acetic acid rather than hydrochloric acid.

Strontium and Pb isotopic ratios were analyzed using a 9-collector, Micromass Sector 54 thermal ionization mass spectrometer. Total procedural blanks are 35 pg for Sr and 60 pg for Pb. Strontium isotopic ratios were fractionation-corrected to 86Sr/88Sr = 0.1194 and are reported relative to 87Sr/86Sr = 0.710254 + 0.000018 (n = 22) for NBS 987. Lead isotopic ratios were analyzed using the double-spike method to correct for mass fractionation during analysis; separate measurements of spiked and unspiked samples were made on different aliquots from the same dissolution. The SBL-74 207Pb–204Pb double-spike from the University of Southampton was used, producing the following results for NBS981: 206Pb/204Pb = 19.9282 ± 0.0023, 207Pb/204Pb = 15.4870 ± 0.0031 and 208Pb/204Pb = 36.6952 ± 0.0097 (n = 7). These values and those of the samples were adjusted to 206Pb/204Pb = 19.9356, 207Pb/204Pb = 15.4891 ± 0.0031 and 208Pb/204Pb = 37.7006 ref.59. 2σ precisions for individual Sr and Pb measurements are better than the analytical errors for NBS987 and NBS981.