Metasomatic Evolution of Coesite-Bearing Diamondiferous Eclogite from the Udachnaya Kimberlite

A coesite-bearing diamondiferous eclogite from the Udachnaya kimberlite (Daldyn field, Siberian craton) has been studied to trace its complex evolution recorded in rock-forming and minor mineral constituents. The eclogite sample is composed of rock-forming omphacite (60 vol%), garnet (35 vol%) and quartz/coesite (5 vol%) and contains intergranular euhedral zoned olivine crystals, up to 200 μm long, coexisting with phlogopite, orthopyroxene, clinopyroxene (secondary), K-feldspar, plagioclase, spinel, sodalite and djerfisherite. Garnet grains are zoned, with a relatively homogeneous core and a more magnesian overgrowth rim. The rim zones further differ from the core in having higher Zr/Y (6 times that in the cores), ascribed to interaction with, or precipitation from, a kimberlite-related melt. Judging by pressure-temperature estimates (~1200 ◦C; 6.2 GPa), the xenolith originated at depths of ~180–200 km at the base of the continental lithosphere. The spatial coexistence of olivine, orthopyroxene and coesite/quartz with K-Na-Cl minerals in the xenolith indicates that eclogite reacted with a deep-seated kimberlite melt. However, Fe-rich olivine, orthopyroxene and low-pressure minerals (sodalite and djerfisherite) likely result from metasomatic reaction at shallower depths during transport of the eclogite by the erupting kimberlite melt. Our results demonstrate that a mixed eclogitic-peridotitic paragenesis, reported previously from inclusions in diamond, can form by interaction of eclogite and a kimberlite-related melt.


Introduction
The origin of eclogite is commonly attributed to the subduction and recycling of Archaean oceanic crust, based on signatures of low-pressure and low-temperature processes (e.g., non-mantle O isotope ratios, Eu anomalies requiring accumulation or fractionation of plagioclase), sometimes followed by mantle metasomatism during their residence in the cratonic mantle lithosphere [1][2][3]. Therefore, mantle-derived eclogites (i.e., those that last resided in the continental lithospheric mantle before entrainment in the kimberlite) can furnish unique information on the Precambrian history of oceanic crust and processes in the upper mantle. [4,5]. To date, there are two eclogite formation models-subduction, which became the paradigm, and high-pressure magmatic. Within the subduction  Primary clinopyroxene grains are anhedral and likewise heavily fractured, of grey-green color with a bluish hue and 0.1 to 1.6 mm in size, with inclusions of garnet and coesite/quartz ( Figure 1). Primary clinopyroxene is omphacite [81] with its composition (Jd 28-29.5 Di 44.5-45 Hd 12.5-13 ) corresponding to Group B eclogite [82]. Omphacite grains are homogeneous, with up to 0.72-1.11 wt% K 2 O and 0.72-0.73 wt% TiO 2 ( Table 1 and Table S1). Clinopyroxene enclosed in the garnet core is compositionally identical to rock-forming omphacite (= CpxI). Table 1. Representative major-element concentrations (wt%; EPMA analyses) for garnet core-rim (Grt) and omphacite (Omp) in diamondiferous coesite-bearing eclogite xenolith from the Udachnaya kimberlite pipe.
White-grey quartz pseudomorphs after coesite ( Figure 1D and Figure 5A) were found in the intergranular space (0.2 to 1.5 mm) or enclosed in garnet and omphacite (up to 200 µm in garnet). Representative Raman spectra of coesite and quartz shown in Figure 3.
Sulfides (pentlandite, pyrrhotite and djerfisherite) occur in the intergranular space as 0.2-2 mm grains ( Figure 5B). Phlogopite is a secondary mineral occurring as 20 to 40 µm grains in the intergranular space together with the secondary mineral assemblage of CpxII-plagioclase-sodalite. Phlogopite was found enclosed in olivine grains or intergrown with orthopyroxene ( Figure 5E). Euhedral crystals of spinel (MgAl 2 O 4 ) with a significant percentage of the hercynite (#38-50) end-member (FeAl 2 O 4 ) were identified in kelyphitic rims around garnets and in the intergranular space together with olivines. Sodalite (100 µm) was detected at the garnet-symplectite boundary. Chemical analyses of phlogopite, spinel and sodalite are listed in Table S1B.
Diamond aggregates have flattened/elongated habits and are located in the intergranular space between rock-forming garnet and omphacite ( Figure 5F). In general, they occur as numerous randomly oriented cubic-octahedral crystals of variable sizes (up to 2 mm) and belong to variety I (octahedral crystal with flat-faces, sharp edges and corners) in the classification of Orlov [92]. The diamond crystals have numerous negative-oriented triangular etch pits on (111) faces with a distinct laminar structure. 0.46 wt% in core and rim, respectively), MnO (0.41-1.55 wt% and 0.52-0.67 wt%), CaO (0.04-0.36 and 0.19-0.37 wt%) and Al2O3 (0.01-0.02 and 0.03-0.13 wt%) (Table S1), whereas the concentrations of NiO and TiO2 do not exceed 0.03 wt% each. The Raman spectra of individual olivine grains show distinct major peaks at 819 and 850 cm −1 and less pronounced peaks at 536 and 915 cm -1 , which are typical olivine Raman peaks ( Figure 3).   Sulfides (pentlandite, pyrrhotite and djerfisherite) occur in the intergranular space as 0.2-2 mm grains ( Figure 5B). Phlogopite is a secondary mineral occurring as 20 to 40 µm grains in the intergranular space together with the secondary mineral assemblage of CpxII-plagioclase-sodalite. Phlogopite was found enclosed in olivine grains or intergrown with orthopyroxene ( Figure 5E). Euhedral crystals of spinel (MgAl2O4) with a significant percentage of the hercynite (#38-50) endmember (FeAl2O4) were identified in kelyphitic rims around garnets and in the intergranular space together with olivines. Sodalite (100 µm) was detected at the garnet-symplectite boundary. Chemical analyses of phlogopite, spinel and sodalite are listed in Table S1B.
Diamond aggregates have flattened/elongated habits and are located in the intergranular space between rock-forming garnet and omphacite ( Figure 5F). In general, they occur as numerous randomly oriented cubic-octahedral crystals of variable sizes (up to 2 mm) and belong to variety I (octahedral crystal with flat-faces, sharp edges and corners) in the classification of Orlov [92]. The diamond crystals have numerous negative-oriented triangular etch pits on (111) faces with a distinct laminar structure.

Garnet
The

Clinopyroxene
Omphacite rare earth elements (REE) patterns show a humped shape with progressive enrichment from La to Eu followed by linear decrease to heavy REE (HREE) (Figure 8). La N (N denotes normalization to chondrite [93]) ranges from 3 to 3.3, Eu n peaks at~10 and Yb n is within 1 to 1.2. The La n /Pr n ratio is 0.6, Gd n /Lu n varies from 9 to 11.7 and Ce n /Eu n is 0.3. Omphacite contains up to 447-470 ppm Sr and 1.7-2.0 ppm Zr. The contents of transition metals are low, with 116-119 ppm Ni, 49.7-52.6 ppm Zn and 4.7-6.4 ppm Cu (Table S2A). The average in situ strontium isotope ratio ( 87 Sr/ 86 Sr) in omphacite is 0.703630 (±0.000047 1σ, n = 7) and 87 Rb/ 86 Sr is low (0.0013) (Table S2B). Hfn; the Zr/Y is 0.07, with 34 ppm Y. The contents of Nb and Ta, as well as Ba, are mostly below the detection limit. Sr abundances is 3.8 ppm, while that of Ni and V is 27 and 137 ppm, respectively (Table S2A).  [94] are shown for reference as well as the range of troctolite compositions from Pito Deep, Pacific ridge [95] and the compositional field of olivine gabbros from the MARK region, mid-Atlantic ridge [96].
Garnet rims exhibit a progressive decrease from HREE to LREE (Figure 8), and medium REE (MREE) contents substantially lower than in the cores: 3.5 Smn, 4.4 Eun and 5.6 Gdn; Ndn is 1.6; the element ratios are Lun/Gdn = 3, Smn/Lan = 27, Gd/Yb = 0.3, Zr/Y = 0.43. The concentrations of Ba, Ta and Rb are below detection limit. Ni and Sr are 23 and 1 ppm, respectively. Some trace elements show prominent core-rim zoning, with respective element contents in core and rim of 2.3 and 8 ppm Zr, 34 and 19 ppm Y, 3.9 and 1 ppm Sr, and below detection limit and 0.2 ppm Hf; Zr/Y in the rim is 6 times higher than in the core (Table S2A).

Geothermobarometry
In the Uv-537 xenolith, primary rock-forming minerals (garnet, omphacite and coesite) enclosed in one another bear signatures of syngenetic growth at least within the coesite stability field. The Sm n *Gd n )) of 1.1. The Lu n /Gd n and Sm n /La n ratios are 1 and 93, respectively. The chondrite-normalized Zr n /Hf n ratio is <1, with 0.6 Zr n and 0.7 Hf n ; the Zr/Y is 0.07, with 34 ppm Y. The contents of Nb and Ta, as well as Ba, are mostly below the detection limit. Sr abundances is 3.8 ppm, while that of Ni and V is 27 and 137 ppm, respectively (Table S2A).
Garnet rims exhibit a progressive decrease from HREE to LREE (Figure 8), and medium REE (MREE) contents substantially lower than in the cores: 3.5 Sm n , 4.4 Eu n and 5.6 Gd n ; Nd n is 1.6; the element ratios are Lu n /Gd n = 3, Sm n /La n = 27, Gd/Yb = 0.3, Zr/Y = 0.43. The concentrations of Ba, Ta and Rb are below detection limit. Ni and Sr are 23 and 1 ppm, respectively. Some trace elements show prominent core-rim zoning, with respective element contents in core and rim of 2.3 and 8 ppm Zr, 34 and 19 ppm Y, 3.9 and 1 ppm Sr, and below detection limit and 0.2 ppm Hf; Zr/Y in the rim is 6 times higher than in the core (Table S2A).

Geothermobarometry
In the Uv-537 xenolith, primary rock-forming minerals (garnet, omphacite and coesite) enclosed in one another bear signatures of syngenetic growth at least within the coesite stability field. The presence of diamond indicates a high pressure of >4 GPa at 950-1000 • C. The concentrations of K 2 O-TiO 2 in omphacite and Na 2 O-TiO 2 in garnet, which are commonly high in ultrahigh-pressure host rocks [12,[97][98][99][100], are rather high in the Uv-537 sample: up to 1.11 wt% and 0.24 wt%, respectively. According to calculations, omphacite with such K 2 O contents forms at 6 GPa [101]. The P-T conditions of eclogite residence in the cratonic lithosphere were reconstructed using the garnet-clinopyroxene geothermometer of Reference [102] and geobarometer of Reference [103]. The garnet core/primary omphacite yielded temperatures of 1200 ± 50 • C (at 6.2 ± 0.2 GPa), that is, within the diamond stability field (along a geothermal gradient corresponding to a surface heat flux of 40 mW·m −2 , appropriate for the lithospheric mantle under Udachnaya [104]).

Bulk Rock Composition
The bulk rock composition of the Uv-537 sample cannot be measured because of its small size and effects of kimberlite infiltration, but can be reconstructed from the chemical composition weighted by modal ratios of its primary minerals, with removed effects of the kimberlite melt, as is common procedure of mantle eclogite xenoliths [6,105,106]. The percentages of minerals were calculated in a thin section, assuming that the relative percentages are constant throughout the sample. The primary bulk composition of Uv-537 was reconstructed using the observed mineralogy of 35% garnet, 60% clinopyroxene and 5% coesite/quartz, without secondary minerals in the intergranular space (olivine, mica, etc.). The density of individual components was calculated proceeding from their relative percentages in minerals (almandine, andradite, grossular and pyrope in garnet and jadeite, diopside and hedenbergite in clinopyroxene, with almandine 4.32; andradite 3.85; grossular 3.59; pyrope 3.58, jadeite 3.25; diopside 3.25, hedenbergite 3.56 g/cm 3 ).
The reconstructed bulk composition of eclogite is similar to gabbroic cumulates, especially in the contents of MgO and CaO [2] (Table S3). The REE spectrum is flat from Sm to HREE while LREE are notably lower. The moderate positive Eu anomaly (Eu* = 1.1) corresponds to that in mafic gabbroic cumulates from more or less fractionated MORB-type melts [107].

Formation of a Mixed Paragenesis during Metasomatism
A diamond crystal with fluid/mineral inclusions provides a snapshot of its crystallization environment. Coesite and olivine inclusions are considered indicator minerals, respectively, of eclogitic or peridotitic origin of diamond [11]. Coexistence of coesite as an SiO 2 -phase (eclogitic suite) and harzburgitic garnet or olivine (peridotitic suite) inclusions within a single diamond crystal record repeated stages of diamond growth in a changeable chemical environment [22,24,108]. The conditions leading to formation of mixed eclogite-peridotite parageneses of inclusions, known from some kimberlite pipes [17][18][19][20][21][22][23][24], are preserved in the chemically inert diamond host. In contrast, they can be fully erased in the diamond substrate (host rock) by the subsequent evolution and homogenization due to diffusion at high temperatures in the mantle or catalyzed by interaction with fluids or melts.
Diamonds from a bimineralic eclogitic xenolith in the Udachnaya kimberlite pipe studied by Ragozin et al. [24] enclose olivine, calcite and intergrown coesite-clinopyroxene. However, the location of these inclusions with respect to the diamond growth zones remains poorly constrained. Yet, no coesite or olivine were found among the rock-forming and accessory minerals of the host xenolith. According to the mineral chemistry of primary clinopyroxene and garnet, the analyzed eclogite belongs to Group A. The origin of this particular eclogite was attributed [24] to C-O-H metasomatic transformation of peridotite.
Wang [23] revealed four olivine, one sanidine and one coesite inclusions in a diamond crystal from a kimberlite in China and hypothesized that diamond originated in an eclogitic host representing subducted oceanic crust later entrained in a mantle plume, followed by extensive eclogite partial melting at peridotite subsolidus conditions. This made it possible to transport the diamond into the peridotite matrix where it subsequently captured peridotitic mineral inclusions during renewed growth from fluids.
A mixed paragenesis of SiO 2 , olivine and harzburgitic garnet inclusions was also reported from diamond in the River Ranch kimberlite (Zimbabwe) [108] and interpreted [22] as evidence of peridotite carbonation. This reaction was described in [109] as leading to the formation of SiO 2 -undersaturated basic magma compositions, precluding the formation of SiO 2 in peridotite. However, strong physicochemical gradients enabling sequential growth of eclogitic and peridotitic minerals might occur locally in highly carbonated metasomatic veins or in a block of subducted sediments [110].

Protolith Reconstruction and Metamorphic Evolution
The bulk rock composition of the Uv-537 eclogite was reconstructed from the garnet core and omphacite, which correspond to the rock-forming minerals in this rock. According to Jacob [2], the presence of coesite implies a low-pressure protolith, as does the presence of moderate positive Eu anomalies (Eu* = 1.1) requiring plagioclase accumulation. We therefore suggest that Uv-537 eclogite formed by subduction of lower oceanic crust composed of gabbronorite or olivine gabbro. Also consistent with a low-pressure protolith are the flat chondrite-normalized MREE-HREE abundances, which are similar to those of modern oceanic crust formed by garnet-absent partial melting of peridotite at spreading ridges, followed by crystallization of olivine, plagioclase and clinopyroxene [3]. The contents of Ni in the reconstructed bulk compositions (44 ppm) are much lower than in primary mantle melts (≥320 ppm) [111] and lower than in group A eclogite (493-659 ppm) [82], that is, the protolith of Uv-537 may have contained little olivine, although some Ni may be stored in sulfide that was not considered in bulk rock reconstruction of the eclogites. Bearing in mind that the behavior of V during partial melting and differentiation is strongly redox-dependent [112], the elevated V content of 157 ppm in the bulk eclogite may reflect a high proportion of cpx in the protolith. A cumulate protolith containing a high proportion of plagioclase is indicated by the positive Eu anomaly in the reconstructed whole rock as well as the extremely low 87 Rb/ 86 Sr of omphacite (0.0013), which controls both Rb and Sr abundances in the bulk rock. This reflects strong Sr partitioning into accumulating plagioclase and exclusion of highly incompatible Rb.
This cumulate protolith was subsequently subducted and underwent metamorphism. Some eclogites from the Udachnaya kimberlite show LREE depletion and HREE enrichment in the reconstructed bulk compositions, possibly resulting from partial melting [106], presumably during or after subduction. As suggested in [113], partial melting reaches the highest degree in basaltic eclogite derived from upper oceanic crust but is insignificant in eclogites derived from gabbros in the deep oceanic crust. Alternatively, the low LREE/HREE are inherited from the crustal protolith, similar to the REE patterns found in the cumulate section of ophiolites [96]. Indeed, the presence of accessory coesite in Uv-537 restricts the melt fraction to 13% [114], although loss of a lower melt fraction cannot be discounted. The available experimental data show that pressure is a key control of high-K clinopyroxene formation [97]. Thus, the high contents of K 2 O in omphacite and Na 2 O in garnet (0.7-1.1 wt% K 2 O and 0.25 wt% Na 2 O, respectively) indicate a deep-seated origin of their host rocks [12] corresponding to ultra-high pressure conditions, as also suggested by the high calculated pressure of >6 GPa and the presence of diamond.

Formation of Olivine and Associated Secondary Minerals
Olivine in eclogite xenoliths is quite rare and poorly investigated [115,116]. According to an experimental study [117], formation of olivine with coesite, K-feldspar and phlogopite can be maintained by increasing the activity of alkalis (K and Na) during metasomatism. High concentrations of K, Na and Cl in primitive kimberlite melts [55,56] provide stability of olivine with coesite, which is precluded under typical mantle conditions where alkali concentrations are low [117]. Unlike most other kimberlites worldwide, some varieties of Udachnaya-East kimberlite remain unaffected by synand postmagmatic alteration. It contains alkali carbonates, sulfates and chlorides in the groundmass (up to 8 wt% Na 2 O + K 2 O and 6 wt% Cl) and is almost anhydrous (<0.5 wt% H 2 O) [80]. In the eclogite under investigation from non-serpentinized ultra-fresh kimberlite from the Udachnaya-East body, the olivine is never in direct contact with coesite but forms part of a secondary assemblage that is restricted to the intergranular space. In this scenario, the relatively forsterite-poor olivine (Fo 83 in core) precipitated along with orthopyroxene and phlogopite from a melt that originated as derivative from a primary (proto-)kimberlite melt (precursor to the host kimberlite) but was hybridized through interaction with eclogite causing the melt to become more Fe-rich. Indeed, zoned olivines with Fo 88 in the core and thin high-Fe rims have been found in carbonate-silicate polymineralic inclusions in a сlinopyroxene megacryst from Diavik A154N kimberlite pipe (see Figure S2 in [59]). These inclusions were interpreted as a recrystallization product of a hybrid melt formed by a reaction between сlinopyroxene, causing its partial dissolution and a percolated kimberlite melt rich in alkalis and carbon dioxide [59]. Although the evidence for partial melt loss from eclogite in this study is ambiguous, as discussed above, LREE-depletion is permissive of small degrees of partial melting of the primary assemblage which may have contributed to the hybridized melt.
The reactions of olivine formation and replacement of primary mineral assemblages in eclogite can be summarized as follows [117]: Secondary orthopyroxene is of quite rare occurrence in mantle eclogites [74]. Its formation is due to interaction of garnet with K-rich fluids: Olivine grains were observed in secondary mineral assemblages, which include sodalite, spinel, plagioclase, djerfisherite, K-feldspar and CpxII. The reconstructed bulk composition of the Uv-537 eclogite does not have such high contents of Na 2 O, K 2 O and Cl. These elements are typical of primitive melt from the Udachnaya pipe [55,56]. Djerfisherite rims around sulfide grains result from reactions with the K-and Cl-bearing fluid [45]. Experimental studies of djerfisherite stability show the crustal origin of this mineral [118,119] which thus formed late during the interaction of the kimberlite melt with the primary eclogitic mineral assemblages. Sodalite is a high-temperature low-pressure (≤0.8 GPa [120]) phase stable above the solidus in sodic silica-undersaturated magmas enriched in NaCl. Thus olivine-sodalite intergrowths can form at pressures below 0.8 GPa. Aegirine (NaFe 3+ Si 2 O 6 ) crystals occurring in the same secondary mineral assemblage indicate a very high activity of oxygen.
According to Misra et al. [45], symplectite forms by partial melting of omphacite and garnet with involvement of K-Cl-H 2 O-CO 2 -fluids, by the reaction Garnet is commonly less strongly altered than omphacite because of their different reactivity with respect to alkaline fluids [74]. Nevertheless, in this study, omphacite appears to be homogeneous whereas garnet features compositionally distinct rims, which are discussed in the next section.

Formation of Heterogeneous Garnet
Zoned garnets are very rare in eclogitic mantle xenoliths and are rarely reported in the literature [121], unlike those from peridotitic xenoliths [122,123]. Prominent zoning in eclogitic garnets from new locations in Angola with different major-oxide contents in the core and the rim (11.2 vs. 10 wt% CaO, 0.06 vs. 0.32 wt% TiO 2 ; 11.20 vs. 15.4 wt% MgO; and 15.2 vs. 10 wt% FeO, respectively) was most likely produced by entrainment in kimberlite and rapid upward transport [121]. Metasomatized zoned harzburgite garnets from the Wesselton kimberlite (South Africa) provide a good example of inhomogeneous garnets formed in subcontinental lithospheric mantle [123]. Sinuous REE patterns of cores in these garnets result from reactions of a strongly depleted protolith with LREE-rich and HFSE-depleted fluids, such as carbonatite. The zoning in these garnets is from low-Ca harzburgitic cores to rims with lherzolitic Ca-Cr relations. Increase in Ca and decrease in Cr and Mg is accompanied by increases in Zr, Y, Ti and HREE. Proceeding from these results, Griffin et al. [123,124] attributed the zoning in garnets to infiltration of a fluid/melt into the rock shortly before intrusion of the kimberlite melt. Infiltration of a melt with a high Zr/Y ratio is additionally confirmed by Zr/Y variations: 1 to 2 in the core vs. 4 to 5 in the rim of garnet grains [124].
The distinct zoning in eclogitic garnets from Uv-537 is manifested as a core-rim difference in major-and trace-element contents. They are, namely, 6.5 and 5 wt% CaO; 0.7 and 0.3 wt% TiO 2 ; 13 and 18.8 wt% MgO; 16 and 12.6 wt% FeO; 0.25 and 0.1 wt% Na 2 O; 2.4 and 8 ppm Zr, 34 and 19 ppm Y; 3.9 and 1 ppm Sr; below detection and 0.2 ppm Hf, in core and rim, respectively. Moreover, the MREE contents but not LREE contents, in the rims are lower than in the cores, which therefore probably does not reflect partial melting but the nature of the metasomatizing melt. The Zr/Y ratio in the rim is 6.6 times that in the core. Previously [62], zoning was reported also for garnets in sheared peridotite xenoliths from the Udachnaya-East kimberlite: Zr/Y = 2.5 in the core and 8.4 in the rim.
Generally, garnets from mantle-derived rocks cannot remain zoned for extended periods of time at the mantle P-T conditions (equilibrium) corresponding to their origin. It was estimated that zonation in garnets from peridotitic xenoliths persisted from 10 to 10 4 years (T = 1200-1400 • C) [125] or from 25 days to 400 years (T = 1025 ± 25 • C, 4.2 ± 0.5 GPa) [126]. As inferred by Griffin et al. [125], metasomatism may be genetically connected to the kimberlite eruptive events or their immediate precursors. According to Jollands et al. [126], single-stage ascent of a kimberlite magma to the surface may take hours to days, and the metasomatic event may mark a precursory failed kimberlite eruption. A similar scenario was suggested [121] for eclogitic xenoliths but with coupled metasomatism and heating during entrainment and/or transport in the host kimberlite. The bulk of the metasomatic alteration of two diamondiferous eclogite xenoliths from the Udachnaya pipe [45], which consist of spongy-textured clinopyroxene, sodalite, calcite, K-feldspar, djerfisherite and a small amount of K-Ca-Fe-Mg glass, was ascribed to a second metasomatic event that occurred also in the upper mantle but after the xenoliths had been incorporated into the kimberlite.
In light of the results in this and prior studies, we suggest that primitive kimberlite melts percolated through the mantle shortly before the formation of the Udachnaya pipe. Their prolonged interaction with eclogite bodies forming part of the regional mantle lithosphere led to hybridization and is responsible for crystallization of the unusual secondary mineral assemblage in Uv-537 as well as for the formation of compositionally distinct, Mg-richer garnet rims. A regression through the data for experiments containing both garnet and olivine in Reference [127] suggests that the garnet rim with an Mg# of 0.73 would be in equilibrium with olivine with an Mg# of 0.79, which is closer to the measured olivine core than the rim composition. This suggests that the metasomatic melt continued to evolve through local reaction with the eclogite, which has a much lower bulk Mg# than the ultramafic metasomatic melt, probably during ascent, leading to the precipitation of forsterite-poorer olivine rims and a low-pressure secondary mineral assemblage (djerfisherite and sodalite), as discussed above.

Metasomatic Reactions of Primitive Melt with Eclogite
There is additional chemical and isotope evidence for interaction of an alkaline kimberlite-like metasomatic melt with Uv-537 eclogite. For example, although the high K 2 O in omphacite and Na 2 O in garnet from Uv-537 undoubtedly in part reflect a pressure effect, as discussed above, these unusually high contents are suggested to additionally reflect metasomatism. This is supported by Rb-Sr isotope systematics in omphacite: the measured 87 Sr/ 86 Sr ratio (0.7036) is high, between modern Depleted Mantle and Primitive Mantle [128], yet 87 Rb/ 86 Sr is very low (0.0013). At the same time, age dating of mantle eclogites from Udachnaya using various radiogenic isotope systems indicates an Archean protolith age [129], suggesting initial 87 Sr/ 86 Sr <~0.7015, depending on the source [130]. At the very low 87 Rb/ 86 Sr, the measured 87 Sr/ 86 Sr cannot be generated by in situ Rb decay in the eclogite as 87 Sr/ 86 Sr increases by only 0.00006 over 2.9 Ga (Table S2) but is strikingly similar to the lower range of 87 Sr/ 86 Sr determined for the Udachnaya kimberlite (0.7034-0.7051; [131]). It therefore seems possible that the eclogite either experienced Sr addition from or isotopically equilibrated with, an ultramafic melt similar to the host kimberlite.
As indicated above, the Udachnaya primary kimberlite melt and/or kimberlite magma had a low-H 2 O and Cl-rich alkali-carbonatitic composition with high bulk Zr/Y of 6-9 early during its ascent [55]. Moreover, some exceptionally well-preserved varieties of non-serpentinized kimberlite are likewise poor in water but rich in Cl, S and alkalis, in the bulk composition as well as within melt inclusions in rock-forming minerals [54,59,79,80,[132][133][134]. As is true in other cratons, such as the Kaapvaal [123,126] and Slave [63], primary kimberlite melt from the source region apparently percolated through the mantle before the formation of the Udachnaya kimberlite. Such melt infiltration could cause partial melting of the mantle material (in particular eclogite with low solidus temperature), crystallization of mineral rims and formation of new phases. The duration and length scale of activity of precursors to successful kimberlite eruption (so-called 'failed kimberlite melts') remain unclear. For instance, microinclusions in diamond from eclogitic xenoliths in the Udachnaya kimberlite that originated at a depth of 160 km [34] have very high contents of alkalis and chlorine, notable amounts of sulfur and low silica, magnesium and water (molar ratio H 2 O/(H 2 O + CO 2 ) = 0.1), at high bulk Zr/Y of 8. In general, their compositions fall in the middle of the range between low-Mg carbonatitic and saline endmember compositions of microinclusions in fibrous diamonds worldwide. Diamond-forming metasomatism was inferred [34] to shortly precede kimberlite eruption, and the microinclusions to represent saline carbonated fluids circulating at the base of the lithospheric mantle (150-170 km depth). The low amounts of water may support a genetic linkage between the microinclusions in diamonds and the host kimberlites [34]. Note that the derivates of the primary kimberlite melt at the Udachnaya site retained their main chemical features even upon prolonged infiltration and remained rich in alkalis, chlorine and CO 2 but poor in silica, magnesium and water. Moreover, studies of primary crystallized multiphase melt inclusions hosted by large metasomatic ilmenite grains in a polymict mantle xenolith from the Bultfontein kimberlite (Kimberley, South Africa) [53] allowed an inference that such melts can percolate through mantle over long distances. Guliani et al. [53] presented the first direct evidence for the existence of alkali-carbonate melts in shallow lithospheric mantle (~110-115 km), above the diamond stability field. From textural and chemical observations, they inferred that the alkali-carbonate melt was derived from primary or precursor kimberlite magmas [53].
At Udachnaya, in any of the presented scenarios, be it infiltration of derivates from the primary kimberlite melt through eclogitic rocks in situ or interaction of the entrained eclogite with the ascending kimberlite magma, the liquid was apparently rich in alkalis and chlorine but poor in H 2 O, as applies to the primitive kimberlite melt [55] and microinclusions in young diamonds [34]. The presence of sodalite (Na 8 Al 6 Si 6 O 24 Cl 2 ), K-feldspar ((K,Na,Ca)(Al 2 Si 2 or AlSi 3 )O 8 ), mica (K(Mg,Fe) 3 [Si 3 AlO 10 ][OH,F] 2 ) and djerfisherite (K 6 (Fe,Ni,Cu) 25 S 26 Cl) among interstitial phases indicates alkali and Cl enrichment of the liquid that interacted with the eclogite. Note that the presence of minor amounts of mica (<1 vol% of the interstitial space) does not mean, by itself, that the melt would be hydrous and <<1 wt% H 2 O are inferred. The preservation of coesite in the Uv-537 sample indicates quite low concentration of water in the system 'eclogite-metasomatizing melt' because, as shown previously [135], coesite-to-quartz conversion is catalyzed by interaction with hydrous melts/fluids.

Summary and Conclusions
The reported assemblage of garnet, omphacitic clinopyroxene and olivine in a diamond-and coesite-bearing eclogitic xenolith sample from the Udachnaya kimberlite allow to place constraints on the formation of a mixed eclogitic-peridotitic mineral paragenesis (Figure 9) known previously only from inclusions in diamond. The mixed paragenesis could be formed by interaction between eclogite and a kimberlite-related melt.
The main conclusions can be summarized as follows: 1. The reconstructed bulk composition of the diamond-coesite eclogite sample Uv-537 indicates its origin from common oceanic protoliths such as mafic gabbroic cumulates which typically form part of the lower crust in mid-ocean ridges. The unusually high contents of K2O in omphacitic clinopyroxene and Na2O in garnet most likely result as a combination of subsequent transport to Figure 9. Cartoon illustrating the pre-kimberlite magmatism history of xenolith Uv-537 and processes involved in its multi-stage evolution (note that depicted mineral volumes are approximate and not designed to represent fluxes). I: Formation of coesite-bearing eclogite (~2.9 Ga) and ancient metasomatic event; II: interaction of eclogite with a metasomatizing carbonatitic or alkali carbonatitic melt (precursory of kimberlite magmatism) produces Mg-rich cores in olivine (Ol, Fo-forsterite number) and Mg-rich rims on garnet (Grt). Omp-omphacite; Coe-coesite; Opx-orthopyroxene; Dia-diamond; Phl-phlogopite; III: transport of eclogite to the surface by kimberlite melt leading to growth of Fe-rich olivine rims, formation of quartz (Qtz) rim around coesite, symplectite after primary omphacite (CpxII-secondary clinopyroxene, Pl-plagioclase, Kfs-K-feldspar), orthopyroxene (?), phlogopite, spinel (Sp), sodalite (Sod), djerfisherite (Dj) and garnet kelyphitic rims.

Summary and Conclusions
The reported assemblage of garnet, omphacitic clinopyroxene and olivine in a diamond-and coesite-bearing eclogitic xenolith sample from the Udachnaya kimberlite allow to place constraints on the formation of a mixed eclogitic-peridotitic mineral paragenesis (Figure 9) known previously only from inclusions in diamond. The mixed paragenesis could be formed by interaction between eclogite and a kimberlite-related melt.
The main conclusions can be summarized as follows: 1.
The reconstructed bulk composition of the diamond-coesite eclogite sample Uv-537 indicates its origin from common oceanic protoliths such as mafic gabbroic cumulates which typically form part of the lower crust in mid-ocean ridges. The unusually high contents of K 2 O in omphacitic clinopyroxene and Na 2 O in garnet most likely result as a combination of subsequent transport to high pressures (≥6 GPa) and metasomatism of eclogite by an alkaline fluid/melt.

2.
Omphacite from the Uv-537 eclogite has a high measured 87 Sr/ 86 Sr ratio (0.7036) that cannot be produced from the in situ decay of the low amount of Rb present ( 87 Rb/ 86 Sr of 0.0013) even for an assumed protolith age of 2.9 Ga. The 87 Sr/ 86 Sr is, however, similar to the lower end of 87 Sr/ 86 Sr determined for the host kimberlite and therefore suggested to reflect isotopic equilibration with, or Sr addition from, a melt with kimberlite affinity.

3.
Zoning in garnet is ascribed to reaction with a percolated and hybridized kimberlite melt shortly before eruption. The Zr/Y ratio (6 times greater in the rim than in the core of garnet grains) records precipitation of the garnet rim from a kimberlite-related melt having high Zr/Y. As mineral zoning cannot persist for extended timescales at upper mantle temperatures, it must reflect processes immediately preceding kimberlite eruption.

4.
The secondary mineral assemblage is formed by reaction of eclogite with derivates of the primary kimberlite melt. Formation of Mg-rich rims in garnet and Mg-rich cores in olivine results from reaction of alkali-rich kimberlite-related melt with eclogite rock residing in the mantle lithosphere beneath Udachnaya. Fe-rich olivine rims, quartz rim around coesite, symplectite after primary omphacite (secondary clinopyroxene, plagioclase, K-feldspar), djerfisherite, sodalite, possibly orthopyroxene and garnet kelyphitic rims were produced by the latest metasomatic reactions during eclogite transport by the kimberlite melt. 5.
The preservation of coesite, along with very fast ascent of the kimberlite melt, supports very low H 2 O contents in the latter. The water-poor kimberlite melt composition is confirmed by the lack of serpentine in the groundmass of a fresh kimberlite sample from the Udachnaya-East pipe [79,80,132,134].