Origin of eclogites from the Earth's mantle

Historically, the first reason to assume an origin as mantle cumulates or melts came from the fact that these rocks could be diamond bearing (Bonney, 1899), which left no doubt of their derivation from deeper than 150 km (Kennedy & Kennedy, 1976) within the Earth's mantle. Cumulate textures of some eclogites were taken as support for this hypothesis by Hatton, (1978), who dismissed the subduction hypothesis on the grounds that some eclogite xenoliths contain 1 wt % Cr₂O₃ or more, which is much higher than the Cr ₂O₃ contents of the oceanic crust. Since the development of the hypothesis that eclogites originate from subducted oceanic crust, a mantle origin tends to be treated as a default alternative. The reason is essentially priority; it was the favoured model before evidence for the subducted crust origin came to light, but has received little attention since then. Snyder et al., (1997) point out that there was a `total lack of definitive indicators for direct mantle derivation' (p. 106) in studies of `accepted' mantle xenoliths and conclude that `the designation of a xenolith or a rock massif as mantle material is based on a lack of crustal indicators' (p. 106). Rather than to falsely state that there are no definitive indicators, it would be more constructive to compile positive geochemical and petrological lines of evidence for eclogites originating from the Earth's mantle. A beginning has been made by Shervais et al., (1988), whose classification contains one group (Group A) to which these authors assign a mantle origin. High Cr₂O₃ contents are often cited as critical evidence in favour of an origin from undisturbed mantle (e.g. Hatton, 1978; Shervais et al., 1988). However, local high-Cr regions may also result from the metamorphism during subduction of chromite layers within either ophiolitic or anorthositic rocks of low-pressure igneous origin.

Origin of eclogites from subducted oceanic crust

Jacob et al., (1994) were able to demonstrate that diamondiferous eclogites from Udachnaya were derived from subducted oceanic crust of Archaean age. The evidence brought by [delta]¹⁸O ratios of garnet and clinopyroxene from the eclogites is indisputable: a range of [delta]¹⁸O ratios was detected (5·19-7·38%^o) that, although smaller than ranges for other eclogite suites (e.g. Roberts Victor, 2-8%^o; Garlick et al., 1971), is undebatably significantly different from the mantle value of 5·5 ± 0·4%^o (Mattey et al., 1994). Correlations of [delta]¹⁸O ratios with various other parameters in these samples can only be explained if attributed to seawater alteration before subduction, and Jacob et al., (1994) attempted to compile and explain these correlations in order to extrapolate to the composition of the unaltered protolith. Snyder et al., (1997) doubt the correlations with [delta]¹⁸O given by Jacob et al., whereas Taylor et al., (1996) bluntly reject them, stating: `The trends reported by Jacob et al., (1994) do not exist' (italics from original). Snyder et al., (1997) base their argument on the fact that their much larger dataset of 33 samples does not show any of the above correlations that Jacob et al., (1994) reported for their dataset of eight eclogites.

In Fig. 1, the samples of Jacob et al., (1994) are plotted together with a larger coherent dataset (59 samples) for clinopyroxene and garnet from eclogites from several South African kimberlite pipes for which oxygen isotope data were also obtained by laser fluorination, and it can be seen that the trends for FeO vs [delta]¹⁸O ( Fig. 1a, b) are obvious and well defined. It should be noted that although the data in Fig. 1 stem from two different cratons, it is still the same process, i.e. seawater alteration, that causes the trends with oxygen isotopes and makes a comparison valid.

Interestingly, Snyder et al., (1997) report a `broad trend of increasing [delta]¹⁸O with increasing ⁸⁷Sr/⁸⁶Sr values' (p. 107; fig. 8 in Snyder et al., 1995), which clearly contradicts their earlier observation that trends like this do not exist. They take this, together with the narrow range of [delta]¹³C ratios displayed by the Udachnaya eclogites (which is in agreement with the extreme restriction of biogenic reservoirs with low [delta]¹³C in the Archaean) as evidence to co-opt the interpretation of Jacob et al., (1994) that the protoliths of Udachnaya eclogites were Archaean oceanic crust. A long-running misinterpretation by the Tennessee group in this context is that Jacob et al., (1994) claimed that all eclogites were subducted oceanic crust, whereas they merely made a generalization that eclogites were formed by subduction beneath both South African and Siberian platforms at about the same time (evidence from age determinations), so that a global process of eclogite building may have occurred at this time. Jacob et al., (1994) were not concerned with defining an origin for all eclogite xenoliths, because they did not attempt to study them all.

Snyder et al.'s interpretation of their data also contradicts earlier publications of their group (e.g. Snyder et al., 1993) in which they postulated that Udachnaya eclogites were `indicating an early ( \>= 4 Ga) differentiation event, whereby the mantle split into complementary depleted and enriched reservoirs'. However, this has already been disproved by various researchers (Ireland et al., 1994; Jacob et al., 1994; Pearson et al., 1995): first, it was demonstrated that the age of the Udachnaya eclogites is late Archaean using two independent isotopic systems (Pb-Pb and Re-Os) that yielded consistent results [2·7 Ga, Jacob et al., (1994); 2·9 Ga, Pearson et al., (1995)]; and, second, Ireland et al., (1994) showed that minerals in eclogite xenoliths experienced metasomatism, which precludes using their present trace element composition to extrapolate to their genesis in any straightforward way.

Snyder et al., (1997) now postulate a `multiplicity' of origins and claim that four of their samples are `probably mantle derived' (p. 107), based partly on whole-rock oxygen isotope compositions which are between 5·42 and 6·05%^o. However, Fig. 1a,b illustrates very clearly that the two `wings' of eclogites with heavy and with lighter oxygen isotope ratios meet at [delta]¹⁸O ratios which are similar to that of undisturbed mantle. Thus, although the oxygen isotopes seem unchanged, eclogites that plot in this area may still have a recycled origin and belong to the same rock suite. Based on plausibility evaluations we think it unlikely that within a suite of eclogites of proven recycled origin some (four) samples should be derived from unrecycled mantle. If Snyder et al., (1997) choose this interpretation they should be able to produce irrefutable evidence in favour of a mantle origin for these samples, which they have yet to deliver.

MINERALOGY AND MAJOR ELEMENT PHASE COMPOSITIONS

The occurrence of kyanite in many Udachnaya eclogite xenoliths puts important constraints on possible petrogenetic models which are neither noted nor discussed by Snyder et al., (1997). It can be shown from phase equilibrium considerations that kyanite-bearing eclogites cannot stably coexist with peridotitic mantle at pressures in the diamond stability field, but would react with olivine to form pyroxenes + garnet. Thus, if these eclogites are of mantle origin, they could only represent cumulates of fractionated high-pressure mantle melts or their metamorphic equivalents that have been kept isolated from chemical communication with surrounding peridotite. However, a fractionation process by which silica-poor, high-pressure mantle melts lead to compositions capable of solidifying to kyanite-bearing rocks has not yet been described. On the other hand, this feature is perfectly compatible with derivation of the eclogites by metamorphism of Si- and Al-rich mantle-derived melts which crystallized at low pressure, i.e. with a derivation from subducted oceanic crust. This implication of the basic mineralogy of the Udachnaya xenoliths is overlooked by Snyder et al., (1997), even though they cite Ringwood, (1990), whose refertilization mechanism is a consequence of this incompatibility. It invalidates both their concluding speculations about large-scale eclogite-mantle interaction as a reason for the relatively narrow range of Udachnaya oxygen isotope data and their speculations about carbonatite as a possible metasomatic agent in the history of these grospydites.

TRACE ELEMENTS: PARTITIONING AND CAUSES

The in situ trace element analyses of clinopyroxene and garnet in the Udachnaya eclogites presented by Snyder et al., (1997) are impressive data that deserve careful analysis. Unfortunately, the interpretation given for LREE partitioning between garnet and clinopyroxene in a subset of their samples is demonstrably erroneous, and is in conflict with other data presented in the same paper. A sub-group of the Udachnaya eclogites (Group A) is characterized by much higher D_Cpx/Gt for La (about 226) than the rest (about 68), which is in turn relatively high compared with the cited experimental determinations. Snyder et al. propose that a low-percentage partial melting event could be the cause for this variation, where `garnet melts first', to drive up the D<sub>La</sub> value for garnet alone (p. 97). This fails to appreciate that D values are equilibrium partitioning values, so that trace element contents of both melt and residuum are controlled during melting by the proportions of the melting phases (garnet > cpx, and not garnet alone as implied), whereas the intermineral distributions of elements reflect the appropriate equilibrium D values at the conditions of melting. The variation of D values as a function of mineral composition is too small to explain the observed variation. In addition to this fundamental misunderstanding, they state that `both Na and Al increase with increasing LREE depletion' (p. 97) in the clinopyroxenes, which is the exact opposite of the expected result of a partial melting event. In fact, the highest D_La values correspond to some of the highest La-in-clinopyroxene abundances of the dataset, which is wholly inconsistent with their being the residues of partial melting.

Other possible explanations are not discussed, such as that the large range of D_Cpx/Gt values may be apparent values representing trace element disequilibrium. The Group A eclogites bearing the high-La clinopyroxenes are the most refractory in terms of Ca-Mg-Fe, so it appears possible that these eclogites underwent relatively recent metasomatic exchange, following which there was insufficient time for re-equilibration of garnet and clinopyroxene.

INTERNAL AGES AND EQUILIBRIUM CONSIDERATIONS

Using the linear correlation between (Sm/Nd)gt and (Sm/Nd)cpx, Snyder et al., (1997) postulate that garnet and clinopyroxene in the Udachnaya eclogites are in chemical equilibrium. Proof of equilibrium is an important issue for scientists working on the determination of intermineral partition coefficients and relies ideally not only on two trace elements, but on a full set of major and trace element and isotopic data (Hauri et al., 1994). Cases are known of rocks where major elements are in equilibrium but trace elements are not (e.g. Jagoutz, 1988), and care has to be taken when analysing internal `ages' of rocks where chemical equilibrium cannot be proven beyond doubt. Disequilibrium is obvious in cases where internal `ages' yield negative values (so-called `future-chrons', e.g. Günther & Jagoutz, 1995), but positive slopes (as observed in Udachnaya eclogites) do not automatically prove the opposite.

The Udachnaya kimberlite consists of two pipes, the `East pipe' being emplaced shortly after the `West pipe' and crosscutting it (Kinny et al., 1998). Therefore, it appears feasible that the thermal event which accompanied the emplacement of the older pipe (Udachnaya West) disturbed the isotopic systematics of the mantle section which is now found as xenoliths in the younger pipe (Udachnaya East) and caused the cluster of internal ages observed by Snyder et al., (1997). However, from 20 internal Sm-Nd ages compiled in table 9 of Snyder et al., (1997), seven are actually younger than the Udachnaya kimberlite (367 ± 4 Ma, Kinny et al., 1998), and nine if we take the kimberlite `dating' of Snyder et al., (1993) (389 ± 4 Ma), a fact that Snyder et al., (1997) fail to provide an explanation for. To accept these internal ages as meaningful requires that at the time of kimberlite eruption, the genesis of many eclogite xenoliths had yet to take place. In our opinion, it is absolutely meaningless to `date' the Udachnaya kimberlite by calculating a weighted average of all internal ages of xenoliths in the kimberlite, as was done by Snyder et al., (1993). This is especially incomprehensible as more reliable ages exist in the literature (e.g. Davis, 1978; Kinny et al., 1998).

RECONSTRUCTED WHOLE-ROCK CHEMICAL COMPOSITIONS

Snyder et al., (1997) present reconstructed whole-rock chemical compositions for their samples as well as for samples from Jacob et al., (1994). They were calculated by combining the modal mineral proportions with the analyses of the clean minerals or, where a modal estimate was not at hand, as in the case of samples from Jacob et al., (1994), by arbitrarily choosing equal amounts of clinopyroxene and garnet. To recalculate `clean' whole rocks is a permissible method in the field of mantle xenoliths which have been subjected to pervasive metasomatism by kimberlite during ascent because measured whole-rock compositions never represent true xenolith compositions, but rather mixtures of kimberlite andxenolith. To reconstruct `clean' whole rocks is especially beneficial for radiogenic isotopes, as was shown by Zindler & Jagoutz, (1988), as these are easily affected by any sort of metasomatism or alteration. However, the Nd isotopic composition of the recalculated `clean' whole rock is very sensitive to imprecisions in the estimations of modal mineralogy: cpx and garnet both host considerable amounts of both Sm and Nd so that small changes in modal compositions translate into rather large isotopic differences of sometimes more than 10 epsilon units. Interestingly, this has been shown by the Tennessee research group itself (Sobolev et al., 1994): these authors showed that whole-rock Nd isotopic evolution lines of two of the recalculated eclogites from Udachnaya, U86/2 and U25/84 [identical with U86 and U25 from Snyder et al., (1993, , 1997)], are almost parallel, and do not intersect within the lifetime of the Earth [fig. 21 of Sobolev et al., (1994) and fig. 4 of Snyder et al., (1993)]. Sobolev et al., (1994) attributed this to imprecision in reconstruction of the whole rock and then proceeded to modify the modal mineralogy of these two samples, adding `slightly more clinopyroxene' to one and `slightly more garnet' to the other and so obtained an intersection at approximately 2 Ga. This example makes it very obvious just how sensitive isotopic composition is to imprecisions in modal mineralogy, and how easily fictive common reservoirs could result.

Imprecisions arise from the generally large grain size of Udachnaya eclogites (0·2-9 mm, Sobolev et al., 1994) combined with their small sample sizes; Sobolev et al., (1994) reported 2-5 cm with exceptions of 10 cm, whereas one or two of Snyder et al.'s have sizes >50 cm. Jacob et al., (1994) stated that in determination of modal mineralogy on samples as small as 10 cm errors up to 30% can easily occur. This can translate into errors of up to 15 epsilon units in calculated clean whole rocks. It should be noted that reconstruction may not lead to large errors for major and trace element compositions, while at the same time being seriously in error for radiogenic isotope systems.

Although it is clear that data containing errors as large as this are not suitable for isotopic dating, Snyder et al., (1997) use reconstructed whole-rock isotopic compositions to arrive at age information (their fig. 14). This plot shows ¹⁴³Nd/¹⁴⁴Nd vs ¹⁴⁷Sm/¹⁴⁴Nd for recalculated whole rocks, for which they chose to use equal amounts of cpx and garnet for the recalculation of whole-rock isotopic compositions (despite the existence of modal mineral estimates), to compare their data with those of Jacob et al., (1994). Snyder et al. observe that: `most data from all three pipes lie along a linear array that yields an "age" of 1·2 Ga and an initial [epsilon]<sub>Nd</sub> of + 2·6, indicating derivation from a LREE-depleted (mantle?) reservoir'. Although the authors admit that `this array is probably a mixing array', they nevertheless state that it `probably reflects only an average age and isotopic composition of the eclogites'.

First of all, it is doubtful if it is suitable to plot data from eclogites from three different kimberlite pipes together on an isochron plot before all three localities are studied in detail and it is assured that apples are not compared with pears. Second, as outlined above, the uncertainties in the data are much too large to achieve meaningful age information or an initial isotopic composition. If, for example, the original modal compositions are taken instead of those chosen by the authors (50:50) to recalculate the whole rocks, the initial [epsilon]°_(CHUR) changes from -28 to -44, a difference of 16 epsilon units! [Note that [epsilon]°_(CHUR) compares the initial ratio with CHUR at the present time, unlike Snyder et al., (1997), who use [epsilon]^t_(CHUR), which compares both at the time given by the `isochron'.] Third, in our opinion, the Sm-Nd isotopic system shows a mixing relationship, and Snyder et al., (1997) seem to prefer a similar explanation. If this is so, and this array represents mixing, then it is impossible that it also provides age information. Thus, fig. 14 of Snyder et al., (1997) merely shows that there is an array of points in ¹⁴³Nd/¹⁴⁴Nd-¹⁴⁷Sm/¹⁴⁴Nd space, but is otherwise meaningless.

In conclusion, although eclogite xenoliths from kimberlite are very complex, a good case can be made for a subducted crustal origin for the majority of the intensively studied suites. To make definitive distinctions between mantle origin for some eclogite xenoliths and a subducted crustal origin for others requires two lines of evidence which Snyder et al., (1997) fail to provide: first, the publication of a complete set of oxygen isotopic data on ultrapure mineral separates of clinopyroxenes and garnets (ideally measured by laser fluorination to have a directly comparable dataset), so that trends between oxygen isotopes and other chemical parameters can be sufficiently well defined; and, second, the compilation of an updated catalogue of the geochemical and petrological characteristics which are to be expected from cumulates from mantle melts at high pressures.

ACKNOWLEDGEMENTS

D. J. would like to thank Steve Foley for fruitful discussions about various petrological and geochemical aspects and for his usefulness as a native speaker, and Tic-Tac-Toe for inspiration. Reviews by R. Rudnick, Z. Sharp and G. Pearson are gratefully acknowledged.

REFERENCES

Clayton, R. N., Goldsmith, J. R., Karel, V. J., Mayeda, T. K. & Newton, R. C. (1975). Limits on the effect of pressure on isotopic fractionation. Geochimica et Cosmochimica Acta 39, 1197-1201.

Davis, G. L. (1978). Zircons from the mantle. Carnegie Institution of Washington, Yearbook 77, 895-897.

Deines, P. (1989). Stable isotope variations in carbonatites. In: Bell, K. (ed.) Carbonatites: Genesis and Evolution. London: Unwin, pp. 301-359.

Garlick, G. D., MacGregor, I. D. & Vogel, D. E. (1971). Oxygen isotope ratios in eclogites from kimberlites. Science 172, 1025-1027.

Gregory, R. T. & Taylor, H. P. (1981). An oxygen isotope profile in a section of Cretaceous oceanic crust, Samail ophiolite, Oman: evidence for [delta]¹⁸O buffering of the oceans by deep (>5 km) seawater-hydrothermal circulation at mid-ocean ridges. Journal of Geophysical Research 86, 2737-2755.

Günther, M. & Jagoutz, E. (1995). Ages and processes as reported by isotopes of kimberlite derived low temperature lherzolites. Extended Abstracts of the 6th International Kimberlite Conference, 205-207.

Hatton, C. J. (1978). The geochemistry and origin of xenoliths from the Roberts Victor mine. Ph.D. Thesis, University of Cape Town.

Hauri, E. H., Wagner, T. P. & Grove, T. L. (1994). Experimental and natural partitioning of Th, U, Pb and other trace elements between garnet, clinopyroxene and basaltic melts. Chemical Geology 117, 149-166.

Ireland, T. R., Rudnick, R. L. & Spetsius, Z. (1994). Trace elements in diamond inclusions reveal links to Archean granites. Earth and Planetary Science Letters 128, 199-213.

Jagoutz, E. (1988). Nd and Sr systematics in an eclogite xenolith from Tanzania: evidence for frozen mineral equilibria in the continental mantle. Geochimica et Cosmochimica Acta 52, 1285-1293.

Jagoutz, E., Dawson, J. B., Hoernes, S., Spettel, B. & Wänke, H. (1984). Anorthositic oceanic crust in the Archean Earth (abstract). Proceedings of the 15th Lunar and Planetary Science Conference. Journal of Geophysical Research, Supplement 395-396.

Jacob, D., Jagoutz, E., Lowry, D., Mattey, D. & Kudrjavtseva, G. (1994). Diamondiferous eclogites from Siberia: remnants of Archean oceanic crust. Geochimica et Cosmochimica Acta 58, 5191-5207.

Kennedy, C. S. & Kennedy, G. C. (1976). The equilibrium boundary between graphite and diamond. Journal of Geophysical Research 81, 2467-2470.

Kinny, P. D., Griffin, B. J., Heaman, L. M., Brakhvogel, F. F. & Spetsius, Z. V. (1998). Shrimp U-Pb ages of perovskite from Yakutian kimberlites. Russian Geology and Geophysics 38, 97-105.

MacGregor, I. D. & Carter, J. L. (1970). The chemistry of clinopyroxenes and garnets of eclogite and peridotite xenoliths from the Roberts Victor mine, South Africa. Physics of the Earth and Planetary Interiors 3, 391-397.

MacGregor, I. D. & Manton, W. I. (1986). Roberts Victor eclogites: ancient oceanic crust. Journal of Geophysical Research 91, 14063-14079.

Mattey, D., Lowry, D. & MacPherson, C. (1994). Oxygen isotope composition of mantle peridotite. Earth and Planetary Science Letters 128, 231-241.

McCulloch, M. T., Gregory, R. T., Wasserburg, G. J. & Taylor, H. P., Jr (1981). Sm-Nd, Rb-Sr and ¹⁸O/¹⁶O isotopic systematics in an oceanic crustal section: evidence from the Samail ophiolite. Journal of Geophysical Research 86, 2721-2735.

Pearson, D. G., Snyder, G. A., Shirey, S. B., Taylor, L. A., Carlson, R. W. & Sobolev, N. V. (1995). Archaean Re-Os age for Siberian eclogites and constraints on Archaean tectonics. Nature 374, 711-713.

Ringwood, A. E. (1990). Slab-mantle interactions: 3. Petrogenesis of intraplate magmas and structure of the upper mantle. Chemical Geology 82, 187-207.

Shervais, J. W., Taylor, L. A., Lugmair, G. W., Clayton, R. N., Mayeda, T. K. & Korotev, R. (1988). Archean oceanic crust and the evolution of sub-continental mantle: eclogites from Southern Africa. Geological Society of America Bulletin 100, 411-423.

Snyder, G. A., Jerde, E. A., Taylor, L. A., Halliday, A. N., Sobolev ,V. N. & Sobolev, N. V. (1993). Nd and Sr isotopes from diamondiferous eclogites, Udachnaya kimberlite pipe, Yakutia, Siberia: evidence of differentiation in the early Earth. Earth and Planetary Science Letters 118, 91-100.

Snyder, G. A., Taylor, L. A., Jerde, E. A., Clayton, R. N., Mayeda, T. K., Deines, P., Rossman, G. R. & Sobolev, N. V. (1995). Archean mantle heterogeneity and the origin of diamondiferous eclogites, Siberia: evidence from stable isotopes and hydroxyl in garnet. American Mineralogist 80, 799-809.

Snyder, G. A., Taylor, L. A., Crozaz, G., Halliday, A. N., Beard, B. L., Sobolev, V. N. & Sobolev, N. V. (1997). The origins of Yakutian eclogite xenoliths. Journal of Petrology 38, 85-113.

Sobolev, V. N., Taylor, L. A. & Snyder, G. A. (1994). Diamondiferous eclogites from the Udachnaya kimberlite pipe, Yakutia. International Geology Reviews 36, 42-64.

Taylor, L. A., Valley, J. W., Clayton, R. N., Mayeda, T. K., Snyder, G. A., Sobolev, V. N. & Sobolev, N. V. (1996). Oxygen isotopes by laser-heating and conventional techniques: a study of Siberian diamondiferous eclogites. Proceedings of the 30th International Geological Congress, Beijing, p. 359.

Zindler, A. & Jagoutz, E. (1988). Mantle cryptology. Geochimica et Cosmochimica Acta 52, 319-335.

*Corresponding author.