Composition of primary kimberlite melt in a garnet lherzolite mantle source: constraints from melting phase relations in anhydrous Udachnaya-East kimberlite with variable CO2 content at 6.5 GPa

The critical issue in the study of kimberlites, known as principal host rocks of diamonds, is the reconstruction of their primary melt composition, which is poorly constrained due to contamination by xenogenic materials, significant loss of volatiles during eruption, and post-magmatic alteration. It is generally accepted that the last equilibration of primary kimberlite melt with surrounding mantle (garnet lherzolite) occurred beneath cratons at 5–7 GPa (150–230 km depths). However, the subliquidus mineral assemblages obtained in kimberlite melting experiments at mantle pressures differ from lherzolite_, probably owing to unaccounted loss of CO₂. Here we present experiments at 6.5 GPa and 1200–1600 °C on unaltered kimberlite with an addition of 2–22 mol% CO₂ over its natural abundance in the rock (13 mol%), but keeping proportions of other components identical to those in an exceptionally fresh anhydrous kimberlite from Udachnaya-East pipe in Siberia. We found that the partial melt achieves equilibrium with garnet lherzolite at 1500 °C and 19–23 mol% CO₂ in the system. Under these conditions this melt contains (mol%): SiO₂ = 9, FeO = 6–7, MgO = 23–26, CaO = 16, Na₂O = 4, K₂O = 1, and CO₂ = 30–35. We propose, therefore, the alkali-rich carbonatitic composition of primary kimberlite melt and loss of 34–45 mol% (34–46 wt%) CO₂ during ascent of the kimberlite magma to the surface.     

Experimental studies At 5–12 GPa of the Ondermatjie hypabyssal kimberlite

Liquidus and sub-liquidus phase relationships are reported for melts formed from an aphanitic kimberlite composition crystallized at 5–12 GPa and 900–1400 °C. The liquidus phase over the pressure range investigated is forsteritic olivine. This is followed with decreasing temperature by olivine plus garnet as the initial sub-liquidus solid phase assemblage. Supra-solidus assemblages consist of olivine+garnet+clinopyroxene+Mg-ilmenite+liquid at 5–7 GPa or olivine+garnet+clinopyroxene+hematite–ilmenite solid solutions (+/−perovskite)+liquid at 8–12 GPa. Phlogopite forms as a near-solidus phase only at 900 °C and 6 GPa. Orthopyroxene does not form at any temperature and pressure. All garnets formed at 6–7 GPa are Ti-rich almandine–grossular–pyrope solid solutions and not Cr-pyrope, whereas garnets formed above 8 GPa are Ti- and Fe³⁺-rich and have no natural counterparts. Quenched liquids are represented by magnesite at 10–12 GPa and Mg–Ca-carbonates at lower pressures. In addition to forming discrete crystals, Mg-ilmenite and hematite–ilmenite solid solutions occur as lamellar intergrowths that are identical in texture to naturally occurring intergrowths. Mg-ilmenite compositions at 6–7 GPa are similar to those of the natural occurrences, whereas clinopyroxenes are richer in Ca. The effects of graphite versus platinum capsules on the oxygen fugacity of the experimental charges and the composition of the olivine, clinopyroxene, Fe–Ti-oxides and garnets formed are described. These experimental data are interpreted to indicate that kimberlite magmas are unlikely to be formed by very small degrees of partial melting of a simple homogeneous carbonated garnet lherzolite mantle. It is proposed that kimberlite magmas form by extensive partial melting of metasomatized mantle, i.e. mineralogically complex carbonate-bearing veins in a lherzolitic/harzburgitic substrate, and that lamellar ilmenite–clinopyroxene intergrowths represent the products of non-equilibrium growth in kimberlite magma.     

Experimental study of the CaMgSi2O6–CO2 system at 3–8 GPa

The melting relationships in the system CaMgSi₂O₆ (Di)–CO₂ have been studied in the 3–8 GPa pressure range to determine if there is an abrupt decrease in the temperature of the solidus accompanying the stabilization of carbonate as a subsolidus phase. Such a decrease has been observed previously in peridotitic and some eclogitic systems. In contrast, the solidus in the Di–CO₂ system was found to decrease in a gradual fashion from 3 to 8 GPa. This decrease accompanies an evolution in the composition of the melt at the solidus from silicate-rich with minor CO₂ at 3 GPa to carbonatitic at 5.5 GPa, where the carbonation reaction Diopside + CO₂ = Dolomite (Dol) + Coesite (Cst) intersects the solidus. The near-solidus melt remains carbonatitic at higher pressure, consistent with carbonate being the dominant contributor to the melt. Based on previous studies in both eclogitic and peridotitic systems, this conclusion can be extended to more complicated systems: once carbonate is a stable subsolidus phase, it plays a major role in controlling both the temperature of melting and the composition of the melt produced.     

Ar and K partitioning between clinopyroxene and silicate melt to 8 GPa

The relative incompatibility of Ar and K are fundamental parameters in understanding the degassing history of the mantle. Clinopyroxene is the main host for K in most of the upper mantle, playing an important role in controlling the K/Ar ratio of residual mantle and the subsequent time-integrated evolution of ⁴⁰Ar/³⁶Ar ratios. Clinopyroxene also contributes to the bulk Ar partition coefficient that controls the Ar degassing rate during mantle melting. The partitioning of Ar and K between clinopyroxene and quenched silicate melt has been experimentally determined from 1 to 8 GPa for the bulk compositions Ab₈₀Di₂₀ (80 mol% albite-20 mol% diopside) and Ab₂₀Di₈₀ with an ultraviolet laser ablation microprobe (UVLAMP) technique for Ar analysis and the ion microprobe for K. Data for Kr (UVLAMP) and Rb (ion probe) have also been determined to evaluate the role of crystal lattice sites in controlling partitioning. By excluding crystal analyses that show evidence of glass contamination, we find relatively constant Ar partition coefficients (D_Ar) of 2.6 × 10⁻⁴ to 3.9 × 10⁻⁴ for the Ab₈₀Di₂₀ system at pressures from 2 to 8 GPa. In the Ab₂₀Di₈₀ system, D_Ar shows similar low values of 7.0 × 10⁻⁵ and 3.0 × 10⁻⁴ at 1 to 3 GPa. All these values are several orders of magnitude lower than previous measurements on separated crystal-glass pairs.D_K is 10 to 50 times greater than D_Rb for all experiments, and both elements follow parallel trends with increasing pressure, although these trends are significantly different in each system studied. The D_K values for clinopyroxene are at least an order of magnitude greater than D_Ar under all conditions investigated here, but D_Ar appears to show more consistent behavior between the two systems than K or Rb. The partitioning behavior of K and Rb can be explained in terms of combined pressure, temperature, and crystal chemistry effects that result in changes for the size of the clinopyroxene M2 site. In the Ab₂₀Di₈₀ system, where clinopyroxene is diopside rich at all pressures, D_K and D_Rb increase with pressure (and temperature) in an analogous fashion to the well-documented behavior of Na. For the Ab₈₀Di₂₀ system, the jadeite content of the clinopyroxene increases from 22 to 75 mol% with pressure resulting in a contraction of the M2 site. This has the effect of discriminating against the large K⁺ and Rb⁺ ions, thereby countering the effect of increasing pressure. As a consequence D_K and D_Rb do not increase with pressure in this system.In contrast to the alkalis (Na, K, and Rb), D_Kr values are similar to D_Ar despite a large difference in atomic radius. This lack of discrimination (and the constant D_Ar over a range of crystal compositions) is also consistent with incorporation of these heavier noble gases at crystal lattice sites and a predicted consequence of their neutrality or “zero charge.” Combined with published D_Ar values for olivine, our results confirm that magma generation is an efficient mechanism for the removal of Ar from the uppermost 200 km of the mantle, and that K/Ar ratios in the residuum are controlled by the amount of clinopyroxene. Generally, Ar is more compatible than K during mantle melting because D_Ar for olivine is similar to D_K for clinopyroxene. As a result, residual mantle that has experienced variable amounts of melt extraction may show considerable variability in time-integrated ³⁶Ar/⁴⁰Ar.     

Experimental study on liquid immiscibility of lamprophyre-sulfide melt at high temperature and high pressure and its geological significance

With lamprophyre and pyrite from the Laowangzhai gold deposit, Yunnan Province, China, as starting materials, and at pressures from 1.5 to 3.0 GPa and temperatures from 1160 to 1560℃ , an experimental study was carried out on the liquid immiscibility of lamprophyre-sulfide melt at high temperature and ultra-high pressure in the DS-29A cubic 3600T/6-type high pressure apparatus. Results showed that the liquid immiscibility of lampro-pyre-sulfide melt in the magmatic system would happen during the early magma...     

Mg-ilmenite megacrysts from the Arkhangelsk kimberlites, Russia: Genesis and interaction with kimberlite melt and postkimberlite fluid

In the present work we studied Mg-ilmenite megacrysts from the Arkhangelsk kimberlites (the Kepino kimberlite field and mantle xenoliths from the Grib pipe). On the basis of isotopic (Rb/Sr, Sm/Nd, δ¹⁸O) and trace-element data we argue that studied Mg-ilmenite megacrysts have a genetic relation to the “protokimberlitic” magma, which was parental to the host kimberlites. Rb-Sr ages measured on phlogopite from ilmenite-clinopyroxenite xenoliths and the host Grib kimberlite overlap within the error (384 Ma and 372 ± 8 Ma, respectively; Shevchenko et al., 2004) with our estimation of the Kotuga kimberlite emplacement (378 ± 25 Ma). Sr and Nd isotopic compositions of megacrysts are close to the isotopic composition of host kimberlites (Mg-ilmenites from kimberlites have ⁸⁷Sr/⁸⁶Sr_{(t = 384)} = 0.7050–0.7063, ?Nd_{(t = 384)} = + 1.7, +1.8, ilmenite from ilmenite-garnet clinopyroxenite xenolith has ⁸⁷Sr/⁸⁶St_{(t = 384)} = 0.7049, ?Nd_{(t = 384)} = +3.5). Oxygen isotopic composition of ilmenites (δ¹⁸O = +3.8–+4.5‰) is relatively “light” in comparison with the values for mantle minerals (δ¹⁸O = +5–+6‰). Taking into account ilmenite-melt isotope fractionation, these values of δ¹⁸O indicate that ilmenites could crystallize from the “protokimberlitic” melt. Temperatures and redox conditions during the formation of ilmenite reaction rims were estimated using ilmenite-rutile and titanomagnetite-ilmenite thermo-oxybarometers. New minerals within the rims crystallized at increasing oxygen fugacity and decreasing temperature. Spinels precipitated during the interaction of ilmenite with kimberlitic melt at T = 1000–1100°C and oxygen fugacity $\Delta \log f_{O_2 }$ [QFM] ≈ 1. Rims comprised with rutile and titanomagnetite crystallized at T ≈ 1100°C, $\Delta \log f_{O_2 }$ [NNO] ≈ 4 and T = 600–613°C, $\Delta \log f_{O_2 }$ [QFM] ≈ 3.7, respectively. Rutile lamellae within ilmenite grains from clinopyroxenitic xenolith were formed T ≥ 1000–1100°C and oxygen fugacity $\Delta \log f_{O_2 }$ [NNO] = ?3.7. Since the pressure of clinopyroxene formation from this xenolith was estimated to be 45–53 kbar, redox conditions at 135–212 km depths could be close to $\Delta \log f_{O_2 }$ [NNO] = ?3.7.     

Fractionation of major elements between coexisting H2O-saturated silicate melt and silicate-saturated aqueous fluids in aluminosilicate systems at 1-2 GPa

From experimental data in the systems Na₂O-Al₂O₃-SiO₂-H₂O, K₂O-Al₂O₃-SiO₂-H₂O at 1100°C, and CaO-Al₂O₃-SiO₂-H₂O at 1200°C in the 1-2 GPa pressure range, the solution behavior of the individual oxides in coexisting H₂O-saturated silicate melts and silicate-saturated aqueous fluids appears to be incongruent. Recalculated on an anhydrous basis, in the CaO-Al₂O₃-SiO₂-H₂O system, CaO^fluid/CaO^melt < 1, whereas in the Na₂O-Al₂O₃-SiO₂-H₂O and K₂O-Al₂O₃-SiO₂-H₂O systems, K₂O^fluid/K₂O^melt and Na₂O^fluid/Na₂O^melt both are greater than 1. The aqueous fluids are depleted in alumina relative to silicate melt.In the Na₂O-Al₂O₃-SiO₂-H₂O, K₂O-Al₂O₃-SiO₂-H₂O, and CaO-Al₂O₃-SiO₂-H₂O systems, fluid/melt partition coefficients for the individual oxides range between ∼0.005 and 0.35 depending on oxide, bulk composition and pressure. The alkali partition coefficients are about an order of magnitude higher than that of CaO. Alumina and silica partition coefficient values in the CaO-Al₂O₃-SiO₂-H₂O system are 10-20% of the values for the same oxides in the Na₂O-Al₂O₃-SiO₂-H₂O and K₂O-Al₂O₃-SiO₂-H₂O systems.Positive correlations among individual partition coefficients and oxide concentrations in the aqueous fluids are consistent with complexing in the fluid that involves silicate polymers associated with alkalis and alkaline earths and aluminosilicate complexes where alkalis and alkaline earths may serve to charge-balance Al³⁺, which is, perhaps, in tetrahedral coordination. Alkali aluminosilicate complexes in aqueous fluid appear more stable than Ca-aluminosilicate complexes.     

Experimental study of magmatic melt oxidation by CO2

The process of CO₂ flashing through hydrous albite-hedenbergite melt was experimentally examined at a temperature of 1100°C and a pressure of 2 kbar. Carbon dioxide was generated when the melt interacted with calcite, and wollastonite was the predominant synthesized phase. Mafic components were introduced into the hydrous albite melt via the dissolution of natural hedenbergite. Raman spectroscopic data on bubbles of the fluid phase in the quench glass indicate that the CO₂/H₂O proportions of the bubbles vary. IR spectroscopic data on the glass prove that the water concentration after CO₂ flashing decreased from 5.5 to approximately 3 wt %. The comparison of the composition of the recrystallized clinopyroxene in contact with melt (with and without CO₂ blowing) indicates that CO₂ oxidizes Fe in the melt. The redox effect of CO₂ is quantified by the empirical clinopyroxene tool for metering oxygen fugacity (oxometer), which was calibrated based on experimental data. The oxygen fugacity in our experiments with CO₂ flashing (estimated by the clinopyroxene oxometer) was NNO + (3.0?C3.5). Our estimates with the application of the clinopyroxene oxometer indicate that the maximum oxygen fugacity in the magmatic chambers of Vesuvius and Stromboli volcanoes (which are bubbled with CO₂) is also close to NNO + (3.5 ± 0.5).     

Beta track autoradiography and infrared spectroscopy bearing on the solubility of CO2 in albite melt at 2 GPa and 1450° C

Albite glasses with 0.5–2.0 wt.% CO₂ synthesized at 2 GPa and 1450° C, previously analyzed by beta track autoradiography (Tingle 1987), have been analyzed by infrared spectroscopy. Values of CO₂ concentration determined by the beta track technique are 10–50% greater than the amount of CO₂ added to the sample. Values determined by infrared spectroscopy match the loaded CO₂ contents, substantiating the existence of a previously unrecognized error in the beta track technique. Concentrations, and hence solubilities, in albite determined by infrared spectroscopy are more accurate than those determined by beta track autoradiography. Values obtained by the beta track technique for diopside are 10% lower than the loaded CO₂ contents (Tingle 1987) and suggest that the systematic error may be dependent on composition and/or density.     

Experimental study of chloritoid stability at high pressure and various fO2 conditions

The reaction chloritoid (ctd)=almandine (alm)+diaspore+H₂O (CAD) has been reversed using Fe³⁺-free synthetic chloritoid and almandine, under fO₂ conditions of the solid oxygen buffer Fe/FeO (CADWI), and using partially oxidized synthetic minerals under fO₂ conditions of the solid oxygen buffer Ni/NiO (CADNNO). Experiments have been conducted between 550 and 700°C, 25 and 45 kbar. The equilibrium pressure and temperature conditions are strongly dependent on the fO₂ conditions (CADNNO lies some-what 50°C higher than CADWI). This can be explained by a decrease in aH₂O for experiments conducted on the Fe/FeO buffer, and a decrease in actd and aalm (through incorporation of ferric iron preferentially in chloritoid) for experiments conducted on the Ni/NiO buffer. The H₂O activity has been calculated using the MRK equation of state, and the values obtained checked against the shift of the equilibrium diaspore=corundum+H₂O bracketed on the Fe/FeO buffer and under unbuffered fO₂ conditions. For fO₂ buffered by the assemblage Fe/FeO, aH₂O increases with pressure from about 0.85 at 600°C, 12 kbar to about 0.9 at 605°C, 25 kbar and 1 above 28 kbar. For fO₂ buffered by the assemblage Ni/NiO, aH₂O=1. The aH₂O decrease from Ni/NiO to Fe/FeO is, however, too small to be entirely responsible for the temperature shift between CADNNO and CADWI. In consequence, the amount of ferric iron in almandine and chloritoid growing in the CADNNO experiments must be significant and change along the CADNNO, precluding calculation of the thermodynamic properties of chloritoid from this reaction. Our experimental data obtained on the Fe/FeO buffer are combined, using a thermodynamic analysis, with Ganguly's (1969) reversal of the reaction chloritoid=almandine+corundum +H₂O (CAC) on the same oxygen buffer. Experimental brackets are mutually consistent and allow extraction of the thermodynamic parameters H ^o f,ctd and S ^octd. Our thermodynamic data are compared with others, generally calculated using Ganguly's bracketing of CACNNO. The agreement between the different data sets is relatively good at low pressure, but becomes rapidly very poor toward high pressure conditions. Using our thermodynamic data for chloritoid and K_D=(Fe³⁺/Al)ctd/(Fe³⁺/Al)alm estimated from natural assemblages, we have calculated the composition of chloritoid and almandine growing from CADNNO and CACNNO. The Fe³⁺ content in chloritoid and almandine increases with pressure, from less than 0.038 per FeAl₂SiO₅(OH)₂ formula unit at 10 kbar to at least 0.2 per formula unit above 30 kbar. This implies that chloritoid and almandine do contain Fe³⁺ in most natural assemblages. The reliability of our results compared to natural systems and thermodynamic data for Mg-chloritoid is tested by comparing the equilibrium conditions for the reaction chloritoid+quartz=garnet (gt)+kyanite+H₂O (CQGK), calculated for intermediate Fe–Mg chloritoid and garnet compositions, from the system FASH and from the system MASH. For 0.65<(XFe)gt<0.8, CQKG calculated from FASH and MASH overlap for K_D=(Mg/Fe)ctd/(Mg/Fe)gt=2. This is in good agreement with the K_D values reported from chloritoid+garnet+quartz+kyanite natural assemblages.     

深海沉积物与履带相互作用试验研究

深海沉积物与履带相互作用的力学特性是深海资源采运设备设计的重要依据。在国内外专家调查、研究的基础上,以膨润土与水的混合物代替深海沉积物,将履带分解为压陷板和剪切板,在特制的试验槽内开展了沉积物的压力沉陷试验与剪切变形试验,得到了沉积物与履带相互作用的压力-沉陷关系曲线和剪应力-剪切位移关系曲线。基于Bekker沉陷模型,得到了模拟沉积物压力-沉陷模型的各个参数值;同时,发现模拟沉积物剪切变形过程具有明显的阶段性,具体可分为线弹性阶段、应变强化阶段、应变软化阶段和残余应力阶段     

100MPa、850℃和800℃条件下锡在流体与富磷过铝质熔体相间分配的实验研究

利用江西宜春414岩体中的钠长石花岗岩作为实验初始物,制备含不同1〉20s含量（0．27％－7．71％）的实验玻璃,本次实验研究了100MPa、850℃和800℃条件下Sn在流体与富磷过铝质熔体相间的分配。实验结果显示,Sn在流体与熔体相间的分配系数（Dsofluid/mclt）变化于2．10×10^-4－1．36×10^3之间,指示Sn强烈趋向于在富磷过铝质熔体中富集。随体系中P2Os含量从0．27％增至1．91％,Sn在流体与熔体相间的分配系数逐渐增加,当体系中R2O5含量进一步增加时,Sn在两相间的分配系数呈降低的趋势。本次实验结果表明,P可能不是Sn以流体相形式进行搬运的主要络合剂。     

Experimental Investigation on Low-Degree Dehydration Partial Melting of Biotite Gneiss and Phengite-Bearing Eclogite at 2 GPa

The ultrahigh-pressure(UHP) eclogite and gneiss from the Dabie(大别)-Sulu(苏鲁) orogen experienced variable degrees of partial melting during exhumation.We report here dehydration partial melting experiments of biotite gneiss and phengite-bearing eclogite at 2 GPa and 800-950 ℃.Our results show that the partial melting of gneiss is associated with the breakdown of biotite into almandine-rich garnet starting at 900 ℃.About 10% granitic melt can be produced at 950 ℃.In contrast,the partial melting of phengite-bea...     

Trace element partitioning and melt structure: An experimental study at 1 atm pressure

Diopside-melt and forsterite-melt rare earth (REE) and Ni partition coefficients have been determined as a function of bulk compositions of the melt. Available Raman spectroscopic data have been used to determine the structures of the melts coexisting with diopside and forsterite. The compositional dependence of the partition coefficients is then related to the structural changes of the melt.The melts in all experiments have a ratio of nonbridging oxygens to tetrahedral cations ($\text{NBO}\text{T}$) between 1 and 0. The quenched melts consist of structural units that have, on the average, 2 (chain), 1 (sheet) and 0 (three-dimensional network) nonbridging oxygens per tetrahedral cation. The proportions of these structural units in the melts, as well as the overall $\text{NBO}\text{T}$, change as a function of the bulk composition of the melt.It has been found that Ce, Sm, Tm and Ni crystal-liquid partition coefficients ($\text{K}{}^{\mathrm{crystal?liq}}{}_{\mathrm{i}}\text{=}\text{C}{}^{\mathrm{crystal}}{}_{\mathrm{i}}\text{C}{}^{\mathrm{liq}}{}_{\mathrm{i}}$) decrease linearly with increasing $\text{NBO}\text{T}$. The values of the individual REE crystal-liquid trace element partition coefficients have different functional relations to $\text{NBO}\text{T}$, so that the degree of light REE enrichment of the melts would depend on their $\text{NBO}\text{T}$.The solution mechanisms of minor oxides such as CO₂, H₂O, TiO₂, P₂O₅ and Fe₂O₃ in silicate melts are known. These data have been recast as changes of $\text{NBO}\text{T}$ of the melts with regard to the type of oxide and its concentration in the melt. From such data the dependence of crystal-liquid partition coefficients on concentration and type of minor oxide in melt solution has been calculated.     

Experimental Melting of Carbonated Peridotite at 6-10 GPa

Partial melting of magnesite-bearing peridotites was studiedat 6–10 GPa and 1300–1700°C. Experiments wereperformed in a multianvil apparatus using natural mineral mixesas starting material placed into olivine containers and sealedin Pt capsules. Partial melts originated within the peridotitelayer, migrated outside the olivine container and formed poolsof quenched melts along the wall of the Pt capsule. This allowedthe analysis of even small melt fractions. Iron loss was nota problem, because the platinum near the olivine container becamesaturated in Fe as a result of the reaction Fe₂SiO₄^Ol = Fe^Fe–Ptalloy + FeSiO₃^Opx + O₂. This reaction led to a gradual increasein oxygen fugacity within the capsules as expressed, for example,in high Fe³⁺ in garnet. Carbonatitic to kimberlite-like meltswere obtained that coexist with olivine + orthopyroxene + garnet± clinopyroxene ± magnesite depending on P–Tconditions. Kinetic experiments and a comparison of the chemistryof phases occasionally grown within the melt pools with thosein the residual peridotite allowed us to conclude that the meltshad approached equilibrium with peridotite. Melts in equilibriumwith a magnesite-bearing garnet lherzolite are rich in CaO (20–25wt %) at all pressures and show rather low MgO and SiO₂ contents(20 and 10 wt %, respectively). Melts in equilibrium with amagnesite-bearing garnet harzburgite are richer in SiO₂ andMgO. The contents of these oxides increase with temperature,whereas the CaO content becomes lower. Melts from magnesite-freeexperiments are richer in SiO₂, but remain silicocarbonatitic.Partitioning of trace elements between melt and garnet was studiedin several experiments at 6 and 10 GPa. The melts are very richin incompatible elements, including large ion lithophile elements(LILE), Nb, Ta and light rare earth elements. Relative to theresidual peridotite, the melts show no significant depletionin high field strength elements over LILE. We conclude fromthe major and trace element characteristics of our experimentalmelts that primitive kimberlites cannot be a direct productof single-stage melting of an asthenospheric mantle. They rathermust be derived from a previously depleted and re-enriched mantleperidotite. KEY WORDS: multianvil; carbonatite melt; peridotite; kimberlite; element partitioning