Marine Carbonates in the Mantle Source of Oceanic Basalts: Pb Isotopic Constraints

For almost fifty years, geochemists have been interpreting the clues from Pb isotopic ratios concerning mantle composition and evolution separately. The Pb isotopes of ocean island basalts (OIB) indicate that their mantle source is heterogeneous, most likely due to the presence of end-components derived from recycled crust and sediment. Some OIB have unusually high 206Pb/204Pb coming from one of the end-components with a long time-integrated high 238U/204Pb or μ (HIMU). Most OIB and many mid-ocean ridge basalts (MORB) also have high 206Pb/204Pb, indicating a HIMU-like source. Moreover, measured 232Th/238U (κ) for most MORB are lower than those deduced from their 208Pb/204Pb and 206Pb/204Pb. Such high μ and low κ features of oceanic basalts are inconsistent with the known geochemical behavior of U, Pb and Th and temporal evolution of the mantle; these have been respectively termed the 1st and 2nd Pb paradox. Here we show that subducted marine carbonates can be a source for HIMU and a solution to the Pb paradoxes. The results are consistent with the predictions of the marine carbonate recycling hypothesis that posits the Pb isotopes of oceanic basalts indicate a common origin and/or magma generation process.

BSE κ that decreased concurrently with continental extraction 15,16 , an upper mantle BSE κ that abruptly changed to 2.5 at 600 Ma 13 and mantle recycling of U due to the Earth's oxidizing condition since Early Proterozoic 2,11,14,17 .
Quite recently, we hypothesized that some Archaean marine carbonates provide the radiogenic 206 Pb/ 204 Pb, unradiogenic 87 Sr/ 86 Sr, low K 2 O, and high CaO of HIMU, as well as the overall positive correlation between CaO and 206 Pb/ 204 Pb in OIB ref. 18 . Our hypothesis is consistent with the Archaean-formed Δ 33 S anomaly in HIMU OIB ref. 19 and the similar trace element contents of carbonatitic melt inclusions in diamonds and HIMU lavas 20 . We also proposed that the unusually radiogenic Pb isotopes of OIB (Fig. 1A) and 1 st and 2 nd 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 equations 21 . 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 mantle 2,11,14,17 .
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 87 Sr/ 86 Sr of coeval seawater 22,23 . 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 87 Sr/ 86 Sr refs 22,23 . 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 alteration [22][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 Shield 27 and one sample (187) is from the 3.0 to 2.6 Ga Bulawayan Supergroup in the Zimbabwe Craton 28 . 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 206 23 are the same within error despite the difference in the leaching procedure used, and are also similar to the recent estimate of Archaean seawater 87 Sr/ 86 Sr ref. 29 .
The 206 Pb/ 204 Pb and 207 Pb/ 204 Pb of bulk and carbonate fractions of our samples plot with or at the extension of the wide 206 Pb/ 204 P-207 Pb/ 204 Pb array of the carbonate fractions of Manjeri limestones 24 that also occur in the Zimbabwe Craton and of Nsuze limestones 26 and Transvaal dolomites 30 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 limestones 25 as well as of surface (plotted) and very radiogenic drill core (not shown) samples   (184, 185, 1977/43), together with some of the Transvaal dolomites, are more radiogenic than OIB. Finally, the 208 Pb/ 204 Pb for given 206 Pb/ 204 Pb 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 208 Pb/ 204 Pb, clearly indicating the extremely low Th content and, hence, Th/U of oxygenated seawater 18,31 . The systematically high 206 Pb/ 204 Pb plus 207 Pb/ 204 Pb and low 208 Pb/ 204 Pb of the carbonate fractions we analyzed clearly indicate that some Archaean limestones have long time-integrated high μ and low κ refs 18,21 . A possibility is that such limestones precipitated locally in Archaean 'oxygen oases' in shallow-platforms 27,32 . 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 ocean 24 . Alternatively, they may have been formed during localized concentrations or 'whiffs' of oxygen in the Archaean 33 , as proposed for the Nsuze metasediments 34 . 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 206 Pb and 207 Pb in their carbonate fraction, rather than from contamination by a high μ Archaean crust [24][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, respectively 2,3 . Irrespective of sources, however, the end-components almost always mix with a 'common' or focus zone (FOZO) component 2,3,37 . 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   (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 alteration [22][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 rocks 38,39 . 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 lithosphere 40,41 . Moreover, high modes of melting would occur once melting initiates, as limestone is mainly comprised of a single calcite phase 42 . 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 melt 43,44 . Subsequent melting of carbonate-metasomatized ancient upper mantle could produce OIB melts 18,21,41 .
To illustrate the origin of the 206 Pb/ 204 Pb and 207 Pb/ 204 Pb 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 refs 7 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 206 Pb and 207 Pb 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 data 41 . Metasomatism (solid line) of a subducted, evolving upper mantle 7 (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 206 Pb and 207 Pb 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 subduction 2,18 and/or later by mantle sulfides 8 , making the slab susceptible to contamination or spiking by the uranogenic Pb refs 18,21 . On the other hand, the low 87 Sr/ 86 Sr of HIMU OIB is simply that of the Archaean limestone, which had the 87 Sr/ 86 Sr of coeval seawater that, in turn, was being buffered by the upper mantle 22,23 (Fig. 2).
Our model illustrates a similar, idealized origin for the 208 Pb/ 204 Pb 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 crust 31 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 208 Pb/ 204 Pb of HIMU OIB is negligible because it had none to very little thorogenic 208 Pb and, compared to Pb, Th was sequestered much less from the slab during subduction. Thus, the 206 Pb/ 204 Pb-208 Pb/ 204 Pb evolution of HIMU OIB (solid line) is primarily an extension of the upper mantle 7 . This notion is consistent with geochemical data indicating a uniform Th/U ratio for the ancient upper mantle 7 and modern DMM refs 21,45,46 .
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 component 18,21 . On the other hand, the HIMU-like 206 Pb/ 204 Pb and radiogenic 87 Sr/ 86 Sr of the bulk of OIB can be produced through mixing between ancient mantle and carbonated crust plus sediment (e.g., containing ankerite 23 ). 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 1 st Pb paradox 21 .
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 206 Pb/ 204 Pb and 208 Pb/ 204 Pb is also an upper mantle signature 2,3,14,16 . However, the fairly recently decoupled low Th/U of MORB (i.e., 2 nd paradox) 13 comes from the aforementioned carbonatitic melts from the down-going slab 18,49 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 238 U/ 235 U of MORB ref. 17 . The 'exotic' 238 U/ 235 U 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 surface 17 .
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 1 st and 2 nd 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-component 31 . 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 87 Sr/ 86 Sr (>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 plumes 31 . 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 subduction 50,51 , there are very few Archaean carbonate outcrops that posses the putative HIMU composition, and carbonate melts 41 and carbonatites 52,53 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 mantle 54,55 and that carbonates can be preserved to great mantle depths 56 . 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 melt 41,57 . 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:HNO 3 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 Oceanography 58 . 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 HNO 3 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 refs 22,23 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 86 Sr/ 88 Sr = 0.1194 and are reported relative to 87 Sr/ 86 Sr = 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-74 207 Pb-204 Pb double-spike from the University of Southampton was used, producing the following results for NBS981: 206 59 . 2σ precisions for individual Sr and Pb measurements are better than the analytical errors for NBS987 and NBS981.

Data Availability
All data generated or analyzed during this study are included in this published article.