Earth and Planetary Science Letters, 93 (1989) 133-141 133 Elsevier Science Publishers B.V., Amsterdam - Printed in The Netherlands

[51

Mantle eclogites: evidence of igneous fractionation in the mantle

J o s e p h R. S m y t h , F lo r i e A. C a p o r u s c i o * a n d T a m s i n C. M c C o r m i c k

Department of Geological Sciences, University of Colorado, Boulder, CO 80309-0250 (U.S.A.)

Received July 6, 1988; revised version received January 10, 1989

Despite recent geochemical studies advocating an origin of eclogite xenoliths in kimberlite by subduction of altered sea-floor basalts and crustal cumulates, there are numerous examples showing abundant exsolution of garnet and kyanite from clinopyroxene that could only have occurred on cooling from near-solidus temperatures at pressures in excess of 3 GPa. Samples showing extensive exsolution tend to have more calcic garnets and sodic pyroxenes than the average eclogite, but do not form a group that is geochemically distinct in major, minor, or trace elements. Also, composite samples containing both kyanite and bimineralic eclogites demonstrate the co-genetic relation of extremes of composition of the eelogite suite. Accordingly, we present an igneous fractionation model to account for the major and trace element variations observed in these rocks: grospydites and kyanite eclogites are early pyroxene, or pyroxene- plus-corundum, cumulates that have undergone extensive subsolidus exsolution of kyanite and garnet. These were then followed by more Fe-rich eclogites which had pyroxene plus garnet as solidus phases and fractionated towards more magnesian bimineralic eelogites.

1. Introduction

Together with alpine-type peridotite and eclo- gite bodies and abyssal peridotites, xenoliths in kimberlites and alkali basalts provide the only natural samples of rock-types f rom the upper mant le (30-200 km). Of these, the xenolith suite in diamondiferous kimberlites, especially those of Southern Africa, appears to have equilibrated at the greatest depths (pressures up to 6.0 G p a and temperatures of 1050-1250 ° C), and to have been exhumed with the min imum of retrograde reac- tions. This suite of xenoliths is domina ted by olivine-rich rocks (depleted garnet peridotites, lherzolites, and harzburgites) which account for 90% or more of all inclusions. Less than 10% of the suite is made up of eclogites and related kyanite eclogites and grospydites (grossular-pyrox- ene-kyanite rocks), and there are no samples inter- mediate in composi t ion between the eclogites and the olivine-bearing suite. However, if the inclu- sions are representative of the region of the mant le f rom which they were derived, then most of the

* Current address: Dipartimento di Scienze della Terra, Uni- versita Degli Studi di Pavia, 1-27100 Pavia, Italy.

0012-821X/89/$03.50 © 1989 Elsevier Science Publishers B.V.

"fert i le" component , that f rom which basaltic melts could be derived, is in eclogites and not in fertile lherzolites or harzburgites. The eclogites display a range of major, minor, and trace element and isotopic composit ions, and an unders tanding of the causes of these variations is essential for a comprehensive view of the origin and evolution of silicate liquids in the Earth.

The general features of mant le eclogites have been reviewed [1,2]. Two major hypotheses have been advanced to account for chemical variat ion in eclogite samples that appear to have equilibrated within the upper mant le ( temperatures - 1050 ° C and pressures > 3 GPa) : (1) a high-pressure cumulate magmat ic origin, and (2) subduct ion of crustal cumulates and oceanic basalts that had been affected by varying degrees of near-surface alteration pr ior to subduction. Several recent stud- ies [3,4] have favored the latter interpretat ion cit- ing geochemical evidence that the samples may be subducted Archean oceanic crust.

Whatever the ult imate origin, the samples have undergone substantial subsolidus re-equilibration and recrystallization which obscures many lines of pet rographic evidence. Nevertheless, in m a n y sam- pies significant textural evidence in the form of garnet and kyanite exsolution f rom pyroxene is

134

preserved which strongly indicates a high-pressure igneous origin at least for the more calcic com- positions. Further, composite samples spanning large compositional ranges demonstrate the co- genetic relationship of the varieties of eclogite. The objectives of this paper are to describe the petrographic, mineralogical, and geochemical characteristics of the eclogite suite, to review the evidence for a high pressure igneous origin, and to present a model for an igneous fractionation pro- cess to account for the major and minor element variations observed.

2. Major element trends

The composition of coexisting garnet and clinopyroxene from 65 eclogites, kyanite eclogites, and grospydites from Bellsbank, South Africa are displayed in Fig. 1. The figure shows that there is a continuous compositional trend from the grospydites to kyanite eclogites, to bimineralic eclogites, and that the grospydites, either with accessory corundum or with accessory coesite and sanidine, are compositionally continuous with the rest of the group in major elements. The crossing tie line is likely due to the elevated alumina con- tent of the pyroxene equilibrated with kyanite. The eclogites that plot to the right in the figure have low-Ca garnets, and fairly high Mg(Mg + Fe)

A

Gr

C Di

Fig. 1. Atomic ACF plot (A = A 1 - N a - K ; C = Ca; F = F e + Mg) of coexisting garnet (open symbol), clinopyroxene (filled symbol), and kyanite ( = A) for 65 eclogites from the Bob- bejaan mine at Bellsbank [5]. The two samples containing the most calcic garnet also contain corundum.

ratios in both garnet and clinopyroxene. Moving to the left, the garnets increase in Ca and gener- ally also in Fe. The pyroxenes also increase in Ca, but generally also in Na, and lift off the bottom of the ACF diagram with increasing grossular con- tent of the coexisting garnet. It is CaEs (Ca-Eskola, CanA12Si4012 ) and CaTs (Ca-Tschermaks, CaA1A1SiO6) components of the pyroxene that causes them to plot off the bottom of the diagram. Kyanite coexists only with garnets that exceed about 35 mol% grossular and pyroxenes with con- siderable CaTs and CaEs components. The grospydites typically have fairly high-Na pyrox- enes (jadeite> 50 mol%) and commonly have accessory corundum or coesite and occasionally sanidine. Among the South African eclogites, coesite, corundum, and sanidine are common only in grospydites, and do not generally occur with low-Ca garnets in these high-pressure eclogites. However, coesite-, quartz- and sanidine-bearing eclogites of broad range of compositions have been reported from other localities [6].

In major element composition, the bimineralic eclogites are similar to oceanic tholeiites and high-alumina basalts, and typically contain small amounts of olivine in the (CIPW) norm (but, no modal olivine). A coesite grospydite [7] contains about 70 mol% normative plagioclase with about 15% each of olivine and diopside. MacGregor and Manton [3] have postulated that such rocks (i.e. kyanite eclogites and grospydites) may be sub- ducted feldspathic cumulates from an Archean oceanic crust. Whereas major element composi- tions of coesite grospydites may resemble anortho- sites, the corundum grospydites do not correspond to any known crustal rock type, being much too low in K and high in Na to correspond to pelitic sediments. Bulk compositions exceed 28% A1203, well outside the range for altered lavas (spilites) [8]. Also grospydites have been reported with both high (> 0.07 wt%) [7] and low (< 0.07 wt%) [9] K20 contents of pyroxenes.

3. Exsolution

There is substantial evidence of abundant exsolution of garnet from clinopyroxene in these rocks (Fig. 2), and recent experimental studies discussed below suggest that this took place on a decrease in temperature rather than an increase in

135

Fig. 2. A photomicrograph of exsolution texture of garnet from clinopyroxene in an eclogite from Roberts Victor showing approximately 40% volume exsolution.

Fig. 3. A photomicrograph of exsolution of both garnet and kyanite from clinopyroxene in a corundum grospydite from Bellsbank (Bobbejaan Mine).

136

pressure. Harte and Gurney [10] discuss extensive garnet exsolution from pyroxene as evidence of a high-pressure igneous origin. The exsolution is significant not only to the petrogenesis of these rocks, but also because it has a first-order effect on the densities of the rocks, so that it will likely control the convection dynamics of an eclogite-rich region of the mantle [5,9].

Exsolution of both kyanite and garnet (Fig. 3) has been noted by several authors [9,11,12]. X-ray precession studies have shown that the structures are epitaxial [9], and the boundaries have been shown to be coherent with high-resolution trans- mission electron microscopy (HRTEM) [13]. This is strong evidence that the garnet and kyanite exsolved from a precursor pyroxene and that all three phases are not the breakdown products of some other precursor phase such as feldspar. Kyanite has a cation/oxygen ratio of 0.60, whereas pyroxene and garnet both have ratios of 0.67. Therefore kyanite exsolution from pyroxene is evi- dence that the precursor pyroxene contained sub- stantial Ca-Eskola end-member. This end-member is unstable below about 3 GPa [14] and is stabi- lized by high temperature (i.e. near solidus), and anomalously high-A1 pyroxenes have been noted as liquidus phases in some experimental studies [14,15]. Because equilibration conditions of these rocks are 3-6 GPa and 1050-1250 ° C, based on Fe-Mg and A1 distributions between garnet and clinopyroxene [16,17], there is not sufficient pres- sure increase between the minimum stability of the GaEs end-member and the final equilibration pressure of the rock to account for the large volume of garnet and kyanite exsolution observed. The exsolution must therefore be primarily due to a decrease in temperature rather than an increase in pressure. The exsolution is, therefore, strong evidence that these kyanite eclogites and grospy- dites showing exsolution were once near liquidus temperatures at pressures in excess of 3 GPa. The precursor pyroxene has very nearly the same com- position as the entire grospydite [9,18]. The crystal chemistry of these pyroxenes has been explored by McCormick [19].

4. Composite samples

Banded eclogites, samples containing two or more chemically distinct eclogites, have been

A

Gr py

C . F

Fig. 4. An atomic ACF plot of co-existing garnet, pyroxene, and kyanite in two composite eclogites [13]. The figure shows the range of compositions occurring across the two composi- tionally banded eclogites and confirms the co-genetic relation- ship of grospydites and bimineralic eclogites.

described by several authors [13,20,21]. The sig- nificance of these samples is that they firmly establish the co-genetic relationship between grospydites and bimineralic eclogites. McCormick [13] has described two composite eclogites that are compositionally banded samples in which grospydites abut bimineralic eclogites. An ACF plot of the coexisting phases is shown in Fig. 4. This figure shows the chemical continuity of grospydites with other eclogites. The range in py- roxene compositions spans the gap between the Group B and C eclogites defined by Shervais and co-workers and thought to be compositionally dis- tinct [4,22]. The compositional gap seen by these authors may possibly be the result of slightly different pyroxene compositions for those equi- librated with kyanite and those not.

5. Trace element trends

Rare earth analyses of whole rocks and mineral separates from these rocks confirm the chemical continuity of the sequence. Representative whole- rock analyses either obtained directly or by recon- struction from cleaned mineral separates are shown in Fig. 5. The lack of LREE enrichment in the grospydite is evidence that there is no contamina- tion of the samples by kimberlite fluids. The bi- mineralic samples are generally flat, at 5-15 ×

100

10.0

1.0

~"'~'A""" i'".~.~.~A ' ....

• .......... : ........

La Ce Nd Sm Eu Gd Dy Er Yb Lu

Fig. 5. A plot of the rare earth concentrat ion relative to chondri tes for a grospydite ( x ) [23], a bimineralic eclogite ('IF) [23], and a kimberlite (A) [24]. Also shown is the rare earth concentrat ion of a natural clinopyroxene relative to a lava ( + ) [24]. The similarity of the latter pat tern with that of the grospydite is consistent with the proposed model for the origin of grospydites as igneous accumulat ions of a luminous pyrox- ene at high pressures.

chondritic and show little or no Eu anomaly. These rocks resemble MORB in trace as well as major elements. The grospydites, however, are dis- tinctive in that they are enriched (about 10-20 x chondritic) in the middle rare earths, Sm, Eu, and Gd, but depleted in both light and heavy REE. The corundum grospydites are similar to the coesite grospydite which suggests that they result from a similar process. Exsolution features dis- cussed above suggest that this process is accumu- lation of an anomalous, liquidus, hyperaluminous pyroxene, with or without corundum, followed by equilibration to lower temperatures with exsolu- tion of kyanite, garnet, and possibly silica. Middle rare earth enrichments have been noted by several authors in both experimental and natural clino- pyroxenes, and one such determination for a natu- ral clinopyroxene relative to host liquid is plotted in Fig. 5.

6. Oxygen isotope studies

Bimineralic mantle eclogite 8180 values show a range of +2.2 to +8.5 [3,25] which is much greater than that shown by either unaltered MORB or mantle-derived peridotites [26,27], but smaller than that shown by ophiolites. Mantle igneous processes are generally believed to be too high in

137

temperature to give rise to significant fractiona- tion of oxygen isotopes [28]. This has prompted several authors [3,29,30] to suggest that the range of 8180 values requires a low-temperature process so that at least some of these xenoliths represent metamorphosed, hydrothermally altered, seafloor basalts which show a range of + 2.0 to + 20 [29]. The kimberlites, themselves, show a range in 8180 values of + 2 to + 12 [31], but this may be due to post-emplacement deuteric alteration. Many eclo- gites show disequilibrium in oxygen isotopes be- tween primary and apparently high-pressure sec- ondary phases, so there may be some degree of oxygen isotope heterogeneity in the subcontinen- tal upper mantle.

There is an insufficient number of oxygen iso- tope data from grospydites and kyanite eclogites to base any firm conclusions as yet, however it appears that the spread in oxygen isotopic ratios may be too great to be the result of mantle ig- neous processes alone, and a more complex his- tory involving low temperature processes is sug- gested. Such low temperature processes may be either in the derivation of the metasomatic fluids that have affected the eclogites within the mantle or in post-emplacement alteration. Also, oxygen isotopic data on grospydites for both whole-rocks and clean mineral separates, particularly of sec- ondary phases if possible, would help to clarify this situation. We are in the process of obtaining these latter data and some preliminary results have been reported [32]. Recent results of Smyth and Clayton [33] in prediction of oxygen isotope fractionation based on electrostatic site potentials computed from crystal structure offer consider- able hope for clarification of oxygen isotope frac- tionation in the mantle.

7. Radiogenic isotope studies

There have been several studies of radiogenic isotopes in mantle eclogites, but none so far has included kyanite eclogites and grospydites. Man- ton and Tatsumoto [34] investigated the U-Pb and Rb-Sr of Roberts-Victor eclogite whole rocks and concluded that the model ages of the rocks were Archean, but that secondary phases might have affected the results. Kramers [35] also studied these isotope systematics from whole-rocks and carefully prepared mineral separates and con-

138

cluded that the ages were Archean, but that the mineral systems had remained open to these iso- topes during the histories of the rocks. This is not surprising for rocks equilibrated at 1050-1250 o C. Jagoutz et al. [30] and McCullough [36] investi- gated the Sm-Nd systematics of several Roberts- Victor eclogites and also concluded that the ages were Archean at 2.5-3.5 Gy, but that the systems had remained open to these isotopes during their histories. MacGregor and Manton [3] have used the Archean ages to infer that the Roberts-Victor eclogites are subducted Archean oceanic crust. Shervais et al. [4] have found unusual Nd isotope ratios (~Nd > 200) in some bimineralic samples which they infer to be the result of metasomatic alteration at low pressures.

8. Discussion

Two major hypotheses have been advanced to account for the chemical and mineralogical varia- tions observed in these rocks: (1) high-pressure igneous fractionation, and (2) subduction of low pressure cumulates that have been affected by varying degrees of near-surface alteration.

Citing predominantly petrographic and major element compositional evidence, several authors have advocated a high-pressure igneous origin for these samples. MacGregor and Carter [37] and Hatton and Gurney [21] describe textures suggest- ing a high-pressure cumulate origin consistent with high-pressure experimental studies [11]. Harte and Gurney [10], Smyth et al. [9], and Lappin [18] cite exsolution phenomena in clinopyroxenes to sug- gest that some of the more calcic eclogites have recrystallized from near solidus temperatures at pressures in excess of 3.0 GPa. These studies have considered the grospydites and kyanite eclogites to be late stage differentiates of this high-pressure igneous fractionation. If derived by igneous processes at depth, the eclogite series would be evidence of basaltic melt at depths greater than 200 km and would provide clues to the igneous fraction of these melts at upper mantle pressures.

The principal shortcomings of this hypothesis are that it fails to account for the lack of trend in Fe enrichment in the grospydites, and it fails to account for the wide variations in oxygen isotopic ratios which appear to require low-temperature processes.

Alternatively, citing predominantly geochem- ical evidence [30,38,39] or studies of North American kimberlite nodules [40-42], other authors have interpreted some of these rocks as possible subducted relicts of an Archean oceanic crust. MacGregor and Manton [3] have further postulated that their group I and group II eclo- gites from Roberts Victor correspond to sub- ducted oceanic crustal cumulates and altered oce- anic basalts respectively. This interpretation is based on a review of major and trace element compositions of whole-rocks and some mineral separates plus oxygen, Sm-Nd, and Rb-Sr isotope studies. The group I and II classification was originally proposed on the basis of texture with group I being high pressure cumulates and group I being fractionated liquids. Revising the interpreta- tion [3] the two types were found to be geochem- ically distinct, with group I (subducted crustal cumulates) having potassic pyroxenes (> 0.07 wt% K20 ) and 8180 values slightly above the mantle average, whereas group II (subducted altered basalts) have low-K pyroxenes and 8180 values below the mantle average. By this classification the grospydites and kyanite eclogites would corre- spond to subducted feldspathic cumulates (group I).

Some of the grospydites, especially corundum- bearing varieties, do not readily fit into this clas- sification. Helmstaedt et al. [40] suggested that the grospydites may be subducted equivalents of rodingites which are Ca- and Al-rich rocks result- ing from near-surface metasomatic alteration of ultramafic and mafic rocks. However, the major element trends observed with rodingitization ap- pear to be quite distinct from those in going from bimineralic eclogites to grospydites. Rodingiti- zation, as observed in Alpine mafic and ultramafic rocks [43], results in strong enrichment of CaO (up to 28 wt.%) with depletion in alkalis, particularly Na, and reduction of silica (down to 36-40 wt.%). Grospydites, however, typically have CaO con- tents of 12-14 wt.%, are not depleted in Na hav- ing pyroxenes with more than 50 mol% jadeite plus occasional primary sanidine, and silica con- tents of 45-52 wt.%.

The major shortcomings of this second hy- pothesis are that it requires a different origin for corundum grospydites whereas these appear to be genetically related to the sequence, and it fails to

account for the evidence that many of the samples have re-equilibrated from near-solidus tempera- tures at pressures in excess of 3 GPa.

Previous authors advocating a high-pressure ig- neous origin have suggested that the grospydites result from evolved eclogitic liquids. We are here postulating that these rocks may be the result of an early accumulation of liquidus pyroxenes fol- lowed by re-equilibrium to lower temperatures at pressures in excess of 3 GPa. This is based on several lines of evidence. First, the reconstructed pyroxene compositions approach the whole-rock grospydites in major element composition. Sec- ond, experimental studies [15] have found that garnet does not readily nucleate in experimental runs, so that a hyperaluminous pyroxene is the liquidus phase. Third, the high Mg/(Mg + Fe) ratios of the grospydites favor an early, rather than a late, formation of these pyroxenes. Fourth, this model would restrict grospydite compositions to the mantle which would be consistent with there being no crustal equivalents to most of these compositions. Fifth, the peculiar rare earth pat- terns of the grospydites are consistent with an igneous accumulation of pyroxene, and do not require plagioclase accumulation to explain the slight positive Eu anomalies.

The igneous process involved appears to be accumulation of a liquidus aluminous pyroxene with or without corundum. Subsequent re-equi- librium results in exsolution or evolution of sub- stantial amounts of garnet, kyanite, and silica. If corundum was present, the evolved silica reacts with it to form additional kyanite resulting in the residual primary corundum being restricted to in- clusions in kyanite or garnet, but never in grain boundary contact with pyroxene. If corundum were not present, the evolved silica may form coesite, although coesite has not been seen in exsolution relationship to the pyroxene. The early pyroxenes have very high Mg/(Mg + Fe) ratios and the peculiar middle-rare-earth enrichment ob- served in both the corundum and coesite grospy- dites. The liquids fractionate and accumulate mag- nesian, high-alumina pyroxenes until garnet nucleates as a liquidus phase with its much lower Mg/(Mg + Fe) ratio. Thereafter, the pyroxenes become much less aluminous, and subsequent fractionation is much more difficult to define. Because of the large radius of the eight-fold site in

139

garnet, it is likely that the garnet would be more iron rich than the liquid so no clear iron enrich- ment pattern, such as results from removal of olivine or orthopyroxene, would be expected.

The spread in oxygen isotopic ratios appears to require that some low-temperature processes be involved, so it may be that the silicate liquids are the result of melting of hydrous subducted oceanic crust. However, our preliminary results suggest that the primary phases are not in oxygen isotopic equilibrium with the secondary phases that appear to be due to metasomatism in the mantle. We believe that further study of stable isotope frac- tionation and heterogeneity in mantle metasoma- tic systems is required before the origin of the liquids can be constrained.

9. Conclusions

Crystal-chemical evidence of extensive exsolu- tion of garnet and kyanite from clinopyroxene together with experimental evidence of liquidus pyroxene compositions indicate that many grospydites and kyanite eclogites are likely to be the result of igneous accumulations of aluminous pyroxenes at pressures in excess of 3 GPa fol- lowed by equilibration to lower temperatures. Composite samples containing both kyanite eclo- gites and bimineralic eclogites confirm the chem- ical continuity of the fractionation trend and show that these rocks are not genetically distinct. Rare earth data on grospydites are consistent with ex- perimental measurements of liquidus pyroxene- liquid distributions and are consistent with a high-pressure pyroxene cumulate origin for these rocks. Oxygen isotopic studies, however, suggest that low-temperature processes may have been involved in the evolution of the liquids, although additional stable and radiogenic isotopic data on grospydites and kyanite eclogites would help to clarify these preliminary conclusions.

The exsolution and geochemical evidence for a high pressure igneous origin for many of the eclo- gites from South African kimberlites indicates that a simple subduction of Archean oceanic crust followed by solid-state metamorphism is not re- sponsible for much of the chemical variation ob- served in these rocks. For eclogite liquids to have existed at depths of 120-150 km, they must have been derived from much greater depths, or geo-

140

thermal gradients must have been much higher in the past. This implies that eclogitic (i.e. basaltic) compositions may be a significant and geochem- ically distinct component throughout the upper mantle. The spread in oxygen isotopic data, how- ever, suggests that low temperature processes may have played a role in the geochemical evolution of this component of the mantle, possibly through low-temperature alteration and subsequent sub- duction and melting of oceanic crust.

Acknowledgements

This work was supported by NSF grant EAR 87-09224. The authors thank Drs. P. Lawless and J.B. Hawthorne of DeBeers Consolidated Mines, Ltd. and Dr. J.J. Gurney of the Department of Geochemistry of the University of Cape Town for samples of the composite eclogites and Drs. I.D. MacGregor, C.R. Stern and G.L. Farmer and one anonymous reviewer for thoughtful reviews of the manuscript.

References

1 J.B. Dawson, Kimberlites and Their Xenoliths, 252 pp., Springer-Verlag, New York, N.Y., 1980.

2 D.A. Carswell, J.B. Dawson, and F.G.F. Gibb, Equilibrium conditions of upper mantle eclogites: implications for kyanite-bearing and diamondiferous varities, Mineral. Mag. 44, 79-89, 1981.

3 I.D. MacGregor and W.I. Manton, The Roberts Victor eclogites: ancient ocean crust?, J. Geophys. Res. 91, 14063-14079, 1986.

4 J.W. Shervais, L.A. Taylor, G.W. Lugmair, R.N. Clayton, T. Mayeda and R.L. Korotev, Early proterozoic oceanic crust and the evolution of subcontinental mantle: eclogites and related rocks from southern Africa, Geol. Soc. Am. Bull. 100, 411-420, 1988.

5 J.R. Smyth and F.A. Caporuscio, Petrology of a suite of eclogite inclusions from the Bobbejaan kimberlite, II. Primary phase compositions and origin, in: Kimberlites, If. The Mantle and Crust-Mantle Relationships, pp. 121-132, Elsevier, Amsterdam, 1984.

6 D.J. Schultze and H. Helmstaedt, Coesite-sanidine eclogites from kimberlite: products of mantle fractionation or sub- duction?, J. Geol. 96, 435-443, 1988.

7 J.R. Smyth and C.J. Hatton, A coesite-sanidine grospydite from the Roberts-Victor kimberlite, Earth Planet Sei. Lett. 34, 284-290, 1977.

8 W. Narebski, A statistical study of specific petrochemical features of some spilitic rock series, in: Spilites and Spilitic Rocks, G.C. Amstutz, ed., pp. 127-160, Springer-Verlag, New York, N.Y., 1974.

9 J.R. Smyth, T.C. McCormick and F.A. Caporuscio, Pet- rology of a suite of eclogite inclusions from the Bobbejaan kimberlite, I. Two unusual corundum-bearing kyanite eclogites, in: Kimberlites, II. The Mantle and Crust-Mantle Rdationships, pp. 109-119, Elsevier, Amsterdam, 1984.

10 B. Harte and J.J. Gurney, Exsolufion of clinopyroxene and garnet in an eclogite nodule from the Roberts Victor Kimberhte pipe, Phys. Chem. Earth 9, 367-387, 1975.

11 M.J. O'Hara and H.S. Yoder, Formation and fractionation of basaltic magmas at high pressure, Scott. J. Geol. 3, 67-117, 1967.

12 N.V. Sobolev, I.K. Kuznetsova and N.I. Zyuzin, The pet- rology of grospydite xenoliths from the Zagodochnaya kimberlite pipe in Yakutia, J. Petrol. 9, 253-280, 1968.

13 T.C. McCormick, Crystal chemistry and breakdown reac- tions of aluminous mantle-derived omphacites, Disserta- tion, Arizona State University, 1984.

14 T. Gasparik and D.H. Lindsley, Phase equilibria at nigh pressure of pyroxenes containing monovalent and trivalent ions, Rev. Mineral. 7, 309-340, 1980.

15 H. Shimizu, Experimental study on rare-earth element par- titioning in minerals formed at 20 and 30 kb for basaltic systems, Geochem. J. 14, 185-202, 1980.

16 D.J. Ellis and D.H. Green, An experimental study on the effect of Ca upon garnet-clinopyroxene Fe-Mg exchange equilibria, Contrib. Mineral. Petrol. 71, 13-22, 1979.

17 T.C. McCormick and J.R. Smyth, Geothermometry and geobarometry for kyanite eclogites from the Bellsbank and Roberts Victor kimberlites, Int. Mineral. Assoc. Abstr. Progr. 169-170, 1986.

18 M.A. Lappin, The evolution of a grospydite from the Roberts-Victor Mine, South Africa, Contrib. Mineral. Pet- rol. 66, 229-241 1978.

19 T.C. McCormick, Crystal-chemical aspects of non-stoichio- metric pyroxenes, Am. Mineral. 71, 1434-1440, 1986.

20 M.A. Lappin and J.B. Dawson, Two Roberts-Victor cumu- late eclogites and their re-equilibration, Phys. Chem. Earth 9, 351-366, 1975.

21 C.J. Hatton and J.J. Gurney, Igneous fractionation trends in Roberts-Victor eclogites, 2nd Int Kimberlite Conf., Ext. Abstr., 1977.

22 L.A. Taylor, C.R. Neal, J.W. Shervais, R.N. Clayton and R.A. Schmitt, Three types of eclogites in the Bellsbank kimberlite, S.A.: crustal and mantle signatures, Trans. Am. Geophys. Union 68, 1551, 1987.

23 F.A. Caporuscio and J.R. Smyth, Trace element composi- tions of mineral separates from mantle eclogites, Trans. Am. Geophys. Union 67, 1253, 1986.

24 C.C. Schnetzler and J.A. Philpotts, Partition coefficients of rare earths between igneous matrix material and rock-form- ing mineral phenocrysts, II, Geochim. Cosmochim. Acta 34, 331-340, 1970.

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

26 K. Muellenbachs and R.N. Clayton, Oxygen isotope studies of fresh and weathered submarine basalts. Can. J. Earth Sci. 9, 172-184, 1974.

27 R.T. Gregory and H.P. Taylor, An oxygen isotope profile in a section of Cretaceous ocean crust, Samail Opniolite,

Oman: evidence of 8180 buffering of the oceans by deep ( > 5 km) seawater hydrothermal circulation of mid-ocean ridges, J. Geophys. Res. 86, 2737-2742, 1981.

28 K. Muehlenbachs and R.N. Clayton, Oxygen isotope stud- ies of fresh and weathered submarine basalts, Can. J. Earth Sci. 9, 172-184, 1972.

29 J.S. Ongley, A.R. Basu and T.K. Kyser, Oxygen isotopes in coexisting garnets, clinopyroxenes and phlogopites of Ro- berts Victor eclogites: implications for petrogenesis and mantle metasomatism, Earth Planet. Sci. Lett. 83, 80-84, 1987.

30 E. Jagoutz, J.B. Dawson, S. Hoernes, B. Spettel and H. Wanke, Anorthositic crust in the Archean, Lunar Planet. Sci. 15, 395-398, 1984.

31 S.M.F. Sheppard and J.B. Dawson, Hydrogen, carbon, and oxygen isotope studies of megacryst and matrix minerals from Lesothan and South African kimberlites. Phys. Chem. Earth 9, 747-763, 1975.

32 F.A. Caporuscio, K.T. Kyser and J.R. Smyth, Oxygen isotopes in mantle eclogites from South Africa, Trans. Am. Geophys. Union 68, 1551-1552, 1987.

33 J.R. Smyth and R.N. Clayton, Correlation of electrostatic site potentials with oxygen isotope fractionation in silicates, Trans. Am. Geophys. Union 69, 1514, 1988.

34 W.I. Manton and M. Tatsumoto, Some Pb and Sr isotopic measurements on eclogites from the Roberts-Victor mine, South Africa, Earth Planet. Sci. Lett. 10, 217-222, 1971.

35 J.D. Kramers, Lead, uranium, strontium, potassium and rubidium in inclusion-bearing diamonds and mantle-de- rived xenoliths from southern Africa, Earth Planet. Sci. Lett. 42, 58-70, 1979.

36 M.T. McCullough, Sm-Nd systematics in eclogite and garnet peridotite nodules from kimberlites: implications for the

141

early differentiation of the Earth, 4th Int. Kimberlite Conf., Abstr., 1986.

37 I.D. MacGregor and J.L. Carter, The chemistry of clino- pyroxenes and garnets of eclogite and peridotite xenoliths from the Roberts Victor Mine, South Africa, Phys. Earth Planet. Int. 3, 391-397, 1970.

38 V. Sautter and B. Harte, Chemical disequilibrium and diffusion gradients in eclogite xenolith JJG41, 4th Int. Kimberhte Conf., Abstr., 1986.

39 P.C. Ater, D.H. Eggler and M.E. McCallum, Petrology and geochemistry of mantle eclogite xenoliths from Colorado- Wyoming kimberlites: recycled ocean crust?, in: Kimber- lites, II. The Mantle and Crust-Mantle Relationships, pp. 309-318, Elsevier, Amsterdam, 1984.

40 H. Helmstaedt, O.L. Anderson and A.T. Gravasci, Petro- fabric studies in eclogite, spinel websterite, and spinel lherzolite xenoliths from kimberlite-bearing breccia pipes in southeastern Utah and northeastern Arizona, J. Geophys. Res. 77, 4350-4365, 1972. '

41 H. Helmstaedt and D.J. Schultze, Garnet clinopyroxene- chlorite eclogite transition in a xenolith from Moses Rock: further evidence for metamorphosed ophiolites under the Colorado Plateau, in: The Mantle Sample: Inclusions in Kimberlites and Other Volcanics, pp. 357-365, American Geophys. Union, Washington, D.C., 1979.

42 H. Helmstaedt and R. Doig, Eclogite nodules from kimberlite pipes of the Colorado Plateau--samples of sub- ducted Franciscan-type oceanic lithosphere, Phys. Chem. Earth 9, 95-111, 1975.

43 B.W. Evans, V. Tromsdorff and G.G. Goles, Geochemistry of high grade eclogites and metarodingites from the Central Alps, Contrib. Mineral. Petrol. 76, 301, 1981.