Simple models for dynamic melting in an upwelling heterogeneous mantle column: Analytical solutions

Several lines of evidence suggest that the melt generation and segregation regions of the mantle are heterogeneous, consisting of chemically and lithologically distinct domains of variable size and dimension. Partial melting of such heterogeneous mantle source regions gives rise to a diverse range of basaltic magmas. In order to better assess the role of source heterogeneity during mantle melting, we have undertaken a theoretical study of trace element distribution and fractionation during concurrent melting and melt migration in an upwelling, chemically heterogeneous, two-porosity double lithology melting column. Analytical solutions for the abundance of a trace element in the matrix and channel were obtained under the assumptions that the porosity, melt and solid velocities, and solid-melt partition coefficients are constant and uniform. For simplicity, we neglected diffusion and dispersion in the melt. Chemical source heterogeneities of arbitrary size and shape were integrated into the simple melting models by allowing trace element abundance in the source region to vary as a function of time and space. Concurrent melting and melt migration in an upwelling heterogeneous mantle may be approximated as a quasi-steady state problem in which time-dependent concentration patterns produced by melting of heterogeneous source regions are superimposed on a reference steady-state concentration distribution established by melting of the ambient or background mantle. Chromatographic fractionation is especially important for the matrix melt and solid when chemical heterogeneities are involved during melting and melt migration in the mantle, giving rise to significant phase-shift between two incompatible trace elements in the matrix melt and scattered correlations among incompatible trace elements in residual peridotites. Mixing is the chief mass transfer process in the dunite channel where the chromatographic effect is negligible for most of the incompatible trace elements. The lack of chromatographic fractionation among incompatible trace elements and isotopic ratios in MORB suggests either most MORB are channel melts or mixing in magma conduit and chamber is very efficient such that the phase-shift is averaged out during magma transport and storage processes. Advection brings melt produced by smaller-degree of melting in the deeper part of the melting column to the overlying melting region, increasing the incompatible trace element abundance in the matrix and the channel. This advection-induced self-enrichment is especially important when heterogeneous sources are involved and may account for some of the enriched incompatible trace element patterns observed in residual peridotite that were previously interpreted to be a result of mantle metasomatism. Systematic studies of high-resolution spatially correlated mantle samples may help to constrain the melting history and the size and nature of chemical heterogeneities in the mantle.     

U-Th-Pb and Lu-Hf isotopic constraints on the evolution of sub-continental lithospheric mantle, French Massif Central

We have carried out a Pb double-spike and Lu-Hf isotope study of clinopyroxenes from spinel-facies mantle xenoliths entrained in Cenozoic intraplate continental volcanism of the French Massif Central (FMC). U-Th-Pb and Lu-Hf isotope systematics verify the existence of different lithospheric domains beneath the northern and southern FMC. Northern FMC clinopyroxenes have extreme Lu/Hf ratios and ultra-radiogenic Hf (ε_Hf = +39.6 to +2586) that reflect ∼15-25% partial melting in Variscan times (depleted mantle model ages ∼360 Ma). Zr, Hf and Th abundances in these clinopyroxenes are low and unaffected by hydrous/carbonatitic metasomatism that overprinted LILE and light REE abundances and caused decoupling of Lu/Hf-Sm/Nd ratios and Nd-Hf isotopes (ε_Nd = +2.1 to +91.2). Pb isotopes of northern FMC clinopyroxenes are radiogenic (²⁰⁶Pb/²⁰⁴Pb > 19), and typically more so than the host intraplate volcanic rocks. ²³⁸U/²⁰⁴Pb ratios range from 17 to 68, and most samples have distinctively low ²³²Th/²³⁸U (<1) and ²³²Th/²⁰⁴Pb (3-22). Clinopyroxenes from southern FMC lherzolites are generally marked by overall incompatible trace element enrichment including Zr, Hf and Th abundances, and have Pb isotopes that are similar to or less radiogenic than the host volcanic rocks. Hf isotope ratios are less radiogenic (ε_Hf = +5.4 to +41.5) than northern FMC mantle and have been overprinted by silicate-melt-dominated metasomatism that affected this part of FMC mantle. Major element and Lu concentrations of clinopyroxenes from southern FMC harzburgites are broadly similar to northern FMC clinopyroxenes and suggest they experienced similar degrees of melt extraction as northern FMC mantle. ²³⁸U/²⁰⁴Pb (53-111) and ²³²Th/²⁰⁴Pb ratios (157-355) of enriched clinopyroxenes from the southern FMC are extreme and significantly higher than the intraplate volcanic rocks. In summary, mantle peridotites from different parts of the FMC record depletion at ∼360 Ma during Variscan subduction, followed by differing styles of enrichment. Northern FMC mantle was overprinted by a fluid/carbonatitic metasomatic agent that carried elements like U, Pb, Sr and light REE. In contrast, much of the southern FMC mantle was metasomatised by a small-degree partial silicate melt resulting in enrichment of all incompatible trace elements. The extreme mantle ²³⁸U/²⁰⁴Pb (northern and southern FMC), ²³²Th/²³⁸U (northern FMC) and ²³²Th/²⁰⁴Pb ratios (southern FMC), coupled with unremarkable present-day Pb isotope ratios, constrain the timing of enrichment. Mantle metasomatism is a young feature related to melting of the upwelling mantle responsible for Cenozoic FMC volcanism, rather than subduction-related metasomatism intimately associated with mantle depletion during the Variscan orogeny. The varying metasomatic styles relate to pre-existing variations in the thickness of the continental lithospheric lid, which controlled the extent to which upwelling mantle could ascend and melt. In the northern FMC, a thicker and more refractory lithospheric lid (?80 km) only allowed incipient degrees of melting resulting in fluid/carbonatitic metasomatism of the overlying sub-continental lithospheric mantle. The thinner lithospheric lid of the southern FMC (?70 km) allowed larger degrees of melting and resulted in silicate-melt-dominated metasomatism, and also focused the location of the volcanic fields of the FMC above this region.     

Effects of melt percolation on highly siderophile elements and Os isotopes in subcontinental lithospheric mantle: A study of the upper mantle profile beneath Central Europe

The effects of melt percolation on highly siderophile element (HSE) concentrations and Re-Os isotopic systematics of subcontinental lithospheric mantle are examined for a suite of spinel peridotite xenoliths from the 4 Ma Kozákov volcano, Bohemian Massif, Czech Republic. The xenoliths have previously been estimated to originate from depths ranging from ∼32 to 70 km and represent a layered upper mantle profile. Prior petrographic and lithophile trace element data for the xenoliths indicate that they were variably modified via metasomatism resulting from the percolation of basaltic melt derived from the asthenosphere. Chemical and isotopic data suggest that lower sections of the upper mantle profile interacted with melt characterized by a primitive, S-undersaturated composition at high melt/rock ratios. The middle and upper layers of the profile were modified by more evolved melt at moderate to low melt/rock ratios. This profile permits an unusual opportunity to examine the effects of variable melt percolation on HSE abundances and Os isotopes.Most HSE concentrations in the studied rocks are significantly depleted compared to estimates for the primitive upper mantle. The depletions, which are most pronounced for Os, Ir and Ru in the lower sections of the mantle profile, are coupled with strong HSE fractionations (e.g., Os_N/Ir_N ratios ranging from 0.3 to 2.4). Platinum appears to have been removed from some rocks, and enriched in others. This enrichment is coupled with lithophile element evidence for the degree of percolating melt fractionation (i.e., Ce/Tb ratio).Osmium isotopic compositions vary considerably from subchondritic to approximately chondritic (γ_Os at 5 Ma from -6.9 to +2.1). The absence of correlations between ¹⁸⁷Os/¹⁸⁸Os and indicators of fertility, as is common in many lithospheric mantle suites, may suggest significant perturbation of the Os isotopic compositions of some of these rocks, but more likely reflect the normal range of isotopic compositions found in the modern convecting mantle. Osmium isotopic compositions correspondingly yield model Re-depletion (T_RD) ages that range from essentially modern to ∼1.3 Ga.Our data provide evidence for large-scale incompatible behavior of HSE during melt percolation as a result of sulfide dissolution, consistent with observations of prior studies. The degree of incompatibility evidently depended on melt/rock ratios and the degree of S-saturation of the percolating melt. The high Pt contents of some of these rocks suggest that the Pt present in this pervasively metasomatized mantle was controlled by a phase unique to the other HSE. Further, high Os concentrations in several samples suggest deposition of Os in a minority of the samples by melt percolation. In these rocks, the mobilized Os was characterized by similar to the ¹⁸⁷Os/¹⁸⁸Os ratios in the ambient rocks. There is no evidence for either the addition of Os with a strongly depleted isotopic composition, or Os with suprachondritic isotopic composition, as is commonly observed under such circumstances.     

Determining melt productivity of mantle sources from U-Th and U-Pa disequilibria; an example from Pico Island, Azores

We use coupled ²³⁸U-²³⁰Th and ²³⁵U-²³¹Pa disequilibria measurements from Pico Island, Azores to examine the melting behavior of the underlying mantle. U-series disequilibria in young, mafic lavas are dependent on the melting rate of their source, which in most cases is primarily controlled by its melt productivity. Mafic lithologies such as eclogite and pyroxenite have much higher melt productivities than peridotite and so U-series measurements may provide constraints on the mineralogy of the melting mantle. Recent Pico Mountain lavas show limited geochemical variations and a restricted range of U-series disequilibria with (²³⁰Th/²³⁸U) = 1.22-1.25 and (²³¹Pa/²³⁵U) = 1.46-1.50. Using a simple, dynamic melting model of a homogeneous source, these results can be reproduced with melting rates of <1 × 10⁻⁴ kg/m³/a and melt porosities of <0.7% near the onset of melting. For a plausible range of upwelling rates, this implies that the melt productivity is <6%/GPa. This value is consistent with a garnet peridotite source but not with more highly productive mafic lithologies. Given independent evidence for the involvement of mafic lithologies such as recycled oceanic crust in Pico magmagenesis, we suggest a scenario in which initial eclogitic melts are dispersed through melt-rock reaction into a larger volume of surrounding peridotite. Subsequent re-melting of the resultant incompatible element enriched peridotite carries a geochemical signature of the mafic lithologies but not necessarily a record of their high melt productivity.     

Variations in melting dynamics and mantle compositions along the Eastern Volcanic Zone of the Gakkel Ridge: insights from olivine-hosted melt inclusions

We present major element, trace element, and volatile concentrations from 66 naturally glassy, olivine-hosted melt inclusions erupted along the Eastern Volcanic Zone (EVZ) of the ultraslow-spreading Gakkel Ridge. Melt inclusion compositions suggest that there are systematic variations in the mantle source composition and melting dynamics from the eastern to the western end of the EVZ. This includes increasing water contents and highly incompatible trace element concentrations (e.g., Ba and Nb) and decreasing light and middle rare earth element concentrations. Ratios of light to heavy rare earth elements in the easternmost melt inclusions are relatively homogeneous, but become more variable to the west. To determine the source of the geochemical variability observed along the EVZ, we model trace elements associated with mantle melting in one- and two-component systems. We consider four possible mantle sources and a range of melting regime shapes, from a full melting triangle to a vertical melting column centered beneath the ridge axes. The observed geochemical variations can be explained by melting of a heterogeneous mantle source composed of depleted MORB mantle plus a metasomatized mantle, where the proportion of the metasomatized component and the extent of melting increases toward the west. Lower rare earth element concentrations and trace element ratios in the westernmost sites also suggest inefficient melt focusing from the outer edges of the melting region. Our results indicate that despite variations in the size of the melting zone and the composition of the mantle source along the ridge axis, the region over which the melts are pooled back to the ridge axis is relatively constant (~10–20 km), suggesting that there is a limit to the distance melts can be transported from off-axis in ultraslow-spreading environments.     

The behaviour of tungsten during mantle melting revisited with implications for planetary differentiation time scales

Tungsten is a moderately siderophile high-field-strength element that is hydrophile and widely regarded as highly incompatible during mantle melting. In an effort to extend empirical knowledge regarding the behaviour of W during the latter process, we report new high-precision trace element data (W, Th, U, Ba, La, Sm) that represent both terrestrial and planetary reservoirs: MORB (11), abyssal peridotites (8), eucrite basalts (3), and carbonaceous chondrites (8). A full trace element suite is also reported for Cordilleran Permian ophiolite peridotites (12) to better constrain the behaviour of W in the upper mantle. In addition, we report our long-term averages for a number of USGS (BIR-1, BHVO-1, BHVO-2, PCC-1, DTS-1) and GSJ (JA-3, JP-1) standard reference materials, some of which we conclude to be heterogeneous and contaminated with respect to W. The most significant finding of this study is that many of the highly depleted upper mantle peridotites contain far higher W concentrations than expected. In the absence of convincing indications for alteration, re-enrichment or contamination, we propose that the W excess was caused by retention in an Os-Ir alloy phase, whose stability is dependent on fO₂ of the mantle source region. This explanation could help to account for the particularly low W content of N-MORB and implies that the lithophile behaviour of W in basaltic rocks is not an accurate representation of the behaviour in the melt source. These findings then become relevant to the interpretation of W-isotopic data for achondrites, where the fractionation of Hf from W during melting is used to infer the Hf/W of the parent body mantle. This is exemplified by the differentiation chronology of the eucrite parent body (EPB), which has been modeled with a melt source with high Hf/W. By contrast, we explore the alternative scenario with a low mantle Hf/W on the EPB. Using available eucrite literature data, a maximum core segregation age of 1.2 ± 1.2 Myr after the closure of CAIs is calculated with a more prolonged time between core formation and mantle fractionation of ca. 2 Myr. This timeline is consistent with most recent published chronologies of the EPB differentiation based on the ⁵³Mn-⁵³Cr and ²⁶Al-²⁶Mg systems.     

Metasomatism of the Pan-African lithospheric mantle beneath the Damara Belt,Namibia, by the Tristan mantle plume: geochemical evidence from mantle xenoliths

In situ trace element analyses of constituent minerals in mantle xenoliths occurring in an alnöite diatreme and in nephelinite plugs emplaced within the central zone of the Damara Belt have been determined by laser ablation ICP-MS. Primitive mantle-normalized trace element patterns of clinopyroxene and amphibole indicate the presence of both depleted MORB-like mantle and variably enriched mantle beneath this region. Clinopyroxenes showing geochemical depletion have low La/Sm_n ratios (0.02–0.2), whereas those showing variable enrichment have La/Sm_n ranging up to 3.8 and La/Yb_n to 9.1. The most enriched clinopyroxenes coexist with amphibole showing similar REE patterns (La/Sm_n = 1.3–4.1; La/Yb_n = 4.5–9). Primitive mantle-normalized trace element patterns allow further groups to be distinguished amongst the variably enriched clinopyroxenes: one having strong relative depletion in Rb–Ba, Ta–Nb and relative enrichment in Th–U; another with similar characteristics but with additional strong relative depletion in Zr–Hf; and one showing no significant anomalies. Amphiboles show similar normalized trace element patterns to co-existing clinopyroxene. Clinopyroxene and amphiboles showing LREE_N enrichment have high Sr and low Nd isotope ratios compared to clinopyroxene with LREE-depleted patterns. Numerical simulation of melt percolation through the mantle via reactive porous flow is used to show that the chromatographic affect associated with such a melt migration process is able to account for the fractionation seen in La–Ce–Nd in cryptically metasomatized clinopyroxenes in Type 1 xenoliths, where melt–matrix interactions occur near the percolation front, whereas REE patterns in clinopyroxenes proximal to the source of metasomatic melt/fluid match those found in modally metasomatized Type 2 xenoliths. The strong fractionation between Rb–Ba, Th–U and Ta–Nb shown by some cryptically metasomatized xenoliths can be also accounted for by reactive porous flow, provided amphibole crystallizes from the percolating melt/fluid close to its source. The presence of amphibole in vein-like structures in some xenoliths is consistent with this interpretation. The strong depletion in Zr–Hf in clinopyroxene and amphibole in some xenoliths cannot be accounted for by melt migration processes and requires metasomatism by a separate carbonate-rich melt/fluid. When taken together with published isotope data on these same xenoliths, the source of metasomatic enrichment of the previously depleted (MORB-like) sub-Damaran lithospheric mantle is attributed to the upwelling Tristan plume head at the time of continental breakup.     

Highly siderophile element composition of the Earth’s primitive upper mantle: Constraints from new data on peridotite massifs and xenoliths

Osmium, Ru, Ir, Pt, Pd and Re abundances and ¹⁸⁷Os/¹⁸⁸Os data on peridotites were determined using improved analytical techniques in order to precisely constrain the highly siderophile element (HSE) composition of fertile lherzolites and to provide an updated estimate of HSE composition of the primitive upper mantle (PUM). The new data are used to better constrain the origin of the HSE excess in Earth’s mantle. Samples include lherzolite and harzburgite xenoliths from Archean and post-Archean continental lithosphere, peridotites from ultramafic massifs, ophiolites and other samples of oceanic mantle such as abyssal peridotites. Osmium, Ru and Ir abundances in the peridotite data set do not correlate with moderately incompatible melt extraction indicators such as Al₂O₃. Os/Ir is chondritic in most samples, while Ru/Ir, with few exceptions, is ca. 30% higher than in chondrites. Both ratios are constant over a wide range of Al₂O₃ contents, but show stronger scatter in depleted harzburgites. Platinum, Pd and Re abundances, their ratios with Ir, Os and Ru, and the ¹⁸⁷Os/¹⁸⁸Os ratio (a proxy for Re/Os) show positive correlations with Al₂O₃, indicating incompatible behavior of Pt, Pd and Re during mantle melting. The empirical sequence of peridotite-melt partition coefficients of Re, Pd and Pt as derived from peridotites () is consistent with previous data on natural samples. Some harzburgites and depleted lherzolites have been affected by secondary igneous processes such as silicate melt percolation, as indicated by U-shaped patterns of incompatible HSE, high ¹⁸⁷Os/¹⁸⁸Os, and scatter off the correlations defined by incompatible HSE and Al₂O₃. The bulk rock HSE content, chondritic Os/Ir, and chondritic to subchondritic Pt/Ir, Re/Os, Pt/Re and Re/Pd of many lherzolites of the present study are consistent with depletion by melting, and possibly solid state mixing processes in the convecting mantle, involving recycled oceanic lithosphere. Based on fertile lherzolite compositions, we infer that PUM is characterized by a mean Ir abundance of 3.5 ± 0.4 ng/g (or 0.0080 ± 0.0009*CI chondrites), chondritic ratios involving Os, Ir, Pt and Re (Os/Ir_PUM of 1.12 ± 0.09, Pt/Ir_PUM = 2.21 ± 0.21, Re/Os_PUM = 0.090 ± 0.002) and suprachondritic ratios involving Ru and Pd (Ru/Ir_PUM = 2.03 ± 0.12, Pd/Ir_PUM = 2.06 ± 0.31, uncertainties 1σ). The combination of chondritic and modestly suprachondritic HSE ratios of PUM cannot be explained by any single planetary fractionation process. Comparison with HSE patterns of chondrites shows that no known chondrite group perfectly matches the PUM composition. Similar HSE patterns, however, were found in Apollo 17 impact melt rocks from the Serenitatis impact basin [Norman M.D., Bennett V.C., Ryder G., 2002. Targeting the impactors: siderophile element signatures of lunar impact melts from Serenitatis. Earth Planet. Sci. Lett, 217-228.], which represent mixtures of chondritic material, and a component that may be either of meteoritic or indigenous origin. The similarities between the HSE composition of PUM and the bulk composition of lunar breccias establish a connection between the late accretion history of the lunar surface and the HSE composition of the Earth’s mantle. Although late accretion following core formation is still the most viable explanation for the HSE abundances in the Earth’s mantle, the “late veneer” hypothesis may require some modification in light of the unique PUM composition.     

Geochemistry of peridotites from the Yap Trench,Western Pacific: implications for subduction zone mantle evolution

ABSTRACT

This study examines the major and trace elements of peridotites from the Yap Trench in the western Pacific to investigate mantle evolution beneath a subduction zone. Major element results show that the peridotites are low in Al₂O₃ (0.31–0.65 wt.%) and CaO (0.04–0.07 wt.%) contents and high in Mg# (Mg/(Mg+Fe)) (0.91–0.92) and have spinels with Cr# (Cr/(Cr+Al)) higher than 0.6 (0.61–0.73). Trace element results show that the peridotites have extremely low heavy rare earth element (HREE) contents compared with abyssal peridotites but have U-shaped chondrite-normalized rare earth element (REE) patterns. The degree of mantle melting estimated based on the major elements, HREEs, and spinel Cr# range from 19% to 25%, indicating that the Yap Trench peridotites may be residues of melting associated with the presence of water in the mantle source. In addition to light rare earth element (LREE) enrichment, the peridotites are characterized by high contents of highly incompatible elements, positive U and Sr anomalies, negative Ti anomalies, and high Zr/Hf ratios. The correlations between these elements and both the degree of serpentinization and high field strength element (HFSE) contents suggest that fluid alteration alone cannot account for the enrichment of the peridotites and that at least the enrichment of LREEs was likely caused by melt–mantle interaction. Comparison between the peridotites and the depletion trend defined by the primitive mantle (PM) and the depleted mantle (DM) suggests that the Yap Trench mantle was modified by subduction-related melt characterized by high contents of incompatible elements, high Zr/Hf ratios, and low HFSE contents. Hydrous melting may have been enhanced by tectonic erosion of the subducting Caroline Plate with complex tectonic morphostructures at the earliest stages of subduction initiation.     

An alternative model for silica enrichment in the Kaapvaal subcontinental lithospheric mantle

The Mg- and Si-rich nature of the sub-cratonic lithospheric mantle (SCLM) beneath the Kaapvaal Craton indicates extensive melt depletion, followed by a Si-enrichment process. Six highly silica enriched peridotites from Kimberley containing high amounts of orthopyroxene (Opx) or garnet (Grt) that are locally concentrated in clots, were investigated to constrain the timing and nature of the Si-enrichment process. A clinopyroxene-bearing lherzolite containing an Opx-clot was studied to quantify the effects of recent metasomatism on the Si-enriched samples. Minerals from the lherzolite, together with Opx from harzburgites and Opx- and Grt-clots have Hf-Nd isotope ratios at the time of kimberlite eruption, 90 Ma, comparable to group I kimberlites and are close to trace element equilibrium with kimberlitic melts. This implies the xenoliths underwent major interaction with kimberlitic melts close to the time of kimberlite eruption.Harzburgites and mineral clots record equilibration pressures and temperatures of, respectively, between 3.5-4.3 GPa and 930-1060 °C. The garnets in Opx-clots have low Lu/Hf and ε_Hf(t) −15, whereas garnets from Grt-clots have high Lu/Hf and ε_Hf(t) +10. In contrast, Grt from both Grt- and Opx-clots have low Sm/Nd and ε_Nd −10. The whole rock platinum group element (PGE) concentrations are an order of magnitude higher in the Grt-clot than the Opx-clot. Measured ¹⁸⁷Os/¹⁸⁸Os range from 0.1085 to 0.1222. The Grt-clot bearing sample yields Nd-Hf-Os isotope model ages that suggest formation in the Neoproterozoic (∼650 Ma). In contrast, an Opx-clot yields T_RD ages of 2.8 Ga, which is interpreted as the time of formation of the host harzburgite. The Opx-clots and host harzburgites have comparable Lu-Hf isotope systematics that imply Opx growth at ∼1.3 Ga and hence their formation is not related to the Grt-clots.Garnets from Opx- and Grt-clots have elevated high-field strength element (HFSE) concentrations, and lack HFSE depletion relative to other trace elements with comparable degrees of incompatibility in the mantle (La/Nb < 0.5). In addition, calculated melts in equilibrium with Grt have strongly fractionated REE (Nd/Yb > 300) and HREE depletion (Yb_N < 0.1) suggesting equilibration with a hydrous melt that is more HREE depleted than a kimberlitic melt. Previous models that related Si-enrichment to subduction are inconsistent with the lack of HFSE depletion (La/Nb < 0.5). Therefore the favoured model for Opx- and Grt-clot formation is infiltration of a hydrous melt in a within plate geodynamical environment associated with volcanism in the Mid-proterozoic and Neoproterozoic, respectively. This implies that Si-enrichment of the Kaapvaal SCLM may be a consequence of numerous localised magmatic events rather than a single craton-wide process.     

Trace element distribution in peridotite xenoliths from Tok, SE Siberian craton: A record of pervasive, multi-stage metasomatism in shallow refractory mantle

Spinel peridotite xenoliths in alkali basalts at Tok, SE Siberian craton range from fertile lherzolites to harzburgites and wehrlites; olivine-rich (70-84%) rocks are dominant. REE patterns in the lherzolites range from nearly flat for fertile rocks (14-17% cpx) to LREE-enriched; the enrichments are positively correlated with modal olivine, consistent with high-permeability of olivine-rich rocks during melt percolation. Clinopyroxene in olivine-rich Tok peridotites typically has convex-upward trace element patterns (La/Nd_PM < 1 and Nd/Yb_PM ? 1), which we consider as evidence for equilibration with evolved silicate liquids (with higher REE and lower Ti contents than in host basalts). Whole-rock patterns of the olivine-rich xenoliths range from convex-upward to LREE-enriched (La/Nd_PM > 1); the LREE-enrichments are positively correlated with phosphorus abundances and are mainly hosted by accessory phosphates and P-rich cryptocrystalline materials. In addition to apatite, some Tok xenoliths contain whitlockite (an anhydrous, halogen-poor and Na-Mg-rich phosphate), which is common in meteorites and lunar rocks, but has not been reported from any terrestrial mantle samples. Some olivine-rich peridotites have generations of clinopyroxene with distinct abundances of Na, LREE, Sr and Zr. The mineralogical and trace element data indicate that the lithospheric mantle section represented by the xenoliths experienced a large-scale metasomatic event produced by upward migration of mafic silicate melts followed by percolation of low-T, alkali-rich melts and fluids. Chromatographic fractionation and fractional crystallisation of the melts close to the percolation front produced strong LREE-enrichments, which are most common in the uppermost mantle and are related to carbonate- and P₂O₅-rich derivatives of the initial melt. Reversal and gradual retreat of the percolation front during thermal relaxation to ambient geotherm (“retrograde” metasomatism) caused local migration and entrapment of small-volume residual fluids and precipitation of volatile-rich accessory minerals. A distinct metasomatic episode, which mainly produced “anhydrous” late-stage interstitial materials was concomitant with the alkali basaltic magmatism, which brought the xenoliths to the surface.     

Trace element partitioning between carbonatitic melts and mantle transition zone minerals: Implications for the source of carbonatites

The occurrence of CO₂-rich lavas (carbonatites, kimberlites) and carbonate-rich xenoliths provide evidence for the existence of carbonatitic melts in the mantle. To model the chemical composition of such melts in the deep mantle, we experimentally determined partition coefficients for 23 trace elements (including REE, U-Th, HFSE, LILE) between deep mantle minerals and carbonatite liquids at 20 and 25 GPa and 1600 °C. Under these conditions, majoritic garnet and CaSiO₃ perovskite are the main reservoirs for trace elements. This study used both femtosecond LA-ICP-MS and SIMS techniques to measure reliable trace element concentrations. Comparison of the two techniques shows a general agreement, except for Sc and Ba. Our experimentally determined partition coefficients are consistent with the lattice strain model. The data suggest an effect of melt structure on partition coefficients in this pressure range. For instance, strain-free partition coefficient (D₀) for majorite-carbonatite melts do not follow the order of cation valence, , observed for majorite-CO₂-free silicate melts. The newly determined partition coefficients were combined with trace element composition of majoritic garnets found as inclusions in diamond to model trace element patterns of deep-seated carbonatites. The result compares favorably with natural carbonatites. This suggests that carbonatites can originate from the mantle transition zone.     

The role of oxidation-reduction and C-H-O fluids in determining melting conditions and magma compositions in the upper mantle

High pressure experimental studies of the melting of lherzolitic upper mantle in the absence of carbon and hydrogen have shown that the lherzolite solidus has a positive dP/dT and that the percentage melting increases quite rapidly above the solidus. In contrast, the presence of carbon and hydrogen in the mantle results in a region of ‘incipient’ melting at temperatures below the C,H-free solidus. In this region the presence or absence of melt and the composition of the melt are dependent on the amount and nature of volatiles, particularly the CO₂, H₂O, and CH₄ contents of the potential C-H-O fluid. Under conditions of low (IW to IW + 1 log unit atP ∼ 20–35kb), fluids such as CH₄+H₂O and CH₄+H₂ inhibit melting, having a low solubility in silicate melts. Under these conditions, carbon and hydrogen are mobile elements in the upper mantle. At slightly higher oxygen fugacity (IW+2 log units,P∼20–35 kb) fluids in equilibrium with graphite or diamond in peridotite C-H-O are extremely water-rich. Carbon is thus not mobile in the mantle in this range and the melting and phase relations for the upper mantle lherzolite approximate closely to the peridotite-H₂O system. Pargasitic amphibole is stable to solidus temperatures in fertile lherzolite compositions and causes a distinctive peridotite solidus, the ‘dehydration solidus’, with a marked change in slope (a ‘back bend’) at 29–30kb due to instability of pargasite at high pressure. Intersections of geothermal gradients with the peridotite-H₂O solidi define the boundary between lithosphere (subsolidus) and asthenosphere (incipient melt region). This boundary is thus sensitive to changes in [affecting CH₄:H₂O:CO₂ ratios] and to the amount of H₂O and carbon (CO₂, CH₄) present. At higher conditions (IW + 3 log units), CO₂-rich fluids occur at low pressures but there is a marked depression of the solidus at 20–21 kb due to intersection with the carbonation reaction, producing the low temperature solidus for dolomite amphibole lherzolite (T∼925°C, 21 to >31kb). Melting of dolomite (or magnesite) amphibole lherzolite yields primary sodic dolomitic carbonatite melt with low H₂O content, in equilibrium with amphibole garnet lherzolite. The complexity of melting in peridotite-C-H-O provides possible explanations for a wide range of observations on lithosphere/asthenosphere relations, on mantle melt and fluid compositions, and on processes of mantle metasomatism and magma genesis in the upper mantle.     

The effect of oxygen fugacity on the partitioning of Re between crystals and silicate melt during mantle melting

Interpretation of Re-Os isotopic systematics applied to mantle and mantle-derived rocks is currently hindered by the poorly understood behaviour of Re and Os during partial melting. Of particular interest is the incompatibility of Re and how it partitions between melt and the different mantle phases. Here, we study the partitioning behaviour of Re between the common upper mantle minerals (garnet, spinel, clinopyroxene, orthopyroxene, and olivine) and silicate melt under temperature (1275-1450 °C) and pressure (1.5-3.2 GPa) conditions relevant for basaltic magma genesis, over a range of oxygen fugacity (?O₂) large enough (QFM+5.6 to QFM−2.9) to demonstrate the effects of changing the oxidation state of Re from 4+ to 6+. Rhenium crystal/silicate-melt partition coefficients vary by 4-5 orders of magnitude, from moderately compatible to highly incompatible, for pyroxenes, garnet, and spinel as the oxidation state of Re changes from 4+ to 6+, but Re in either oxidation state is incompatible in olivine. Because the changeover from the one Re oxidation state to the other occurs over the range of ?O₂s pertinent to partial melting in the Earth’s mantle, bulk Re crystal/silicate-melt partition coefficients during mantle melting are also expected to vary significantly according to the oxidation state of the system. For instance, assuming QFM−0.7 and QFM+1.6 as average ?O₂ for mid-ocean ridge (MORBs) and island arc (IABs) basalts, respectively, a difference of at least one order of magnitude for bulk Re partition coefficients is expected (excluding any influence from a sulphide phase). Hence, Re is probably much more incompatible during the genesis of IABs compared to MORBs. Our results also demonstrate that Re⁴⁺ has a partitioning behaviour similar to Ti⁴⁺ rather than Yb, and is accordingly not a sensitive indicator of garnet in the source. The lower concentrations of Re observed in ocean island basalts (OIBs) compared to MORBs are therefore not a result of being generated deeper in the mantle where garnet is stable, leaving the hypothesis of late-stage loss of Re from OIB lavas by degassing as the most plausible explanation.     

Geochemical characteristics of the La Réunion mantle plume source inferred from olivine-hosted melt inclusions from the adventive cones of Piton de la Fournaise volcano (La Réunion Island)

Major and trace element compositions were obtained for bulk rocks and melt inclusions hosted in olivine crystals (Fo > 85) from the adventive cones of the Piton de La Fournaise volcano (La Réunion Island). Ratios of highly incompatible trace elements for these magmas are used to identify the nature of the La Réunion mantle plume source. Although adventive cone lavas display unusual major element compositions compared to the historical lavas of the volcano (e.g., lower CaO/Al₂O₃), trace element data suggest that the magmas emitted by the adventive cones originate from a common chemical source. This source may correspond to either a homogeneous mixed source of different mantle components or a near-primitive less-differentiated mantle source. The melt inclusions display ratios of highly incompatible elements (e.g., Th/La, Nb/La) which are similar to primitive mantle values, and lower Nb/U ratios compared to most oceanic basalts. These results and previous isotopic and trace element data suggest that La Réunion plume samples a source which is intermediate between a primitive-like mantle domain and a slightly depleted one almost unaffected by the recycling processes. This source could have originated from early depletion of the primitive mantle. Assuming a depletion 4.45 Gyr ago, ~10% melting of this slightly depleted source could explain the enriched trace element concentrations of the melt inclusions.     

藏北新生代玄武质火山岩起源的深部机制——大陆俯冲和板片断离驱动的地幔对流上涌模式

近年来地震层析成像揭示出可可西里-西昆仑中新世-第四纪钾质火山岩带下方存在一个深达900km的巨型地幔低速体,空间上与新特提斯洋和印度大陆俯冲断离板片沉降形成的冷地幔下降流共存(Replumaz et al.,2010a,b),两者构成统一的地幔对流体系。研究表明,羌塘古近纪(60~34Ma)钠质玄武岩和高钾钙碱性玄武岩均以富含Ti O2、P2O5和大离子亲石元素为特征,主体具有与OIB相近的微量元素组成和弱亏损的Sr、Nd同位素特征,指示岩浆起源于软流圈的上涌熔融,但Nb、Ta的弱亏损表明岩浆源区有岩石圈地幔熔融组分的贡献。羌塘(32~26Ma)碱性钾质玄武岩与可可西里和西昆仑中新世以来喷发的钾质玄武岩的地球化学性质相近,不相容元素比值和Sr、Nd同位素组成指示岩浆起源于古俯冲地幔楔的低程度熔融。这些特征表明藏北软流圈上涌作用始于古近纪,初始上涌中心位于羌塘地体之下。计算表明藏北古近纪火山岩距离当时的印度大陆北缘的最大和最小距离约为1250km和700km,与现今可可西里地幔低速体的南、北边界与印度大陆北缘的距离相近,支持羌塘古近纪地幔上涌作用也是受藏南冷地幔下降流所驱动。青藏高原在南北缩短过程中不仅表现为软流圈自西向东挤出流动,地幔垂向对流也是其重要的运动形式,在地幔上升流形成的藏北热幔区内,地壳的水平缩短增厚与岩石圈地幔的伸展减薄呈脉动式共存。藏南冷地幔下降流和藏北热地幔上升流的持续北移是导致藏北后碰撞火山岩时空迁移的主要控制因素。     

Experimental comparison of trace element partitioning between clinopyroxene and melt in carbonate and silicate systems, and implications for mantle metasomatism

Experiments in the systems diopside-albite (Di-Ab) and diopside-albite-dolomite (Di-Ab-Dmt), doped with a wide range of trace elements, have been used to characterise the difference between clinopyroxene-silicate melt and clinopyroxene-carbonate melt partitioning. Experiments in Di-Ab-Dmt yielded clinopyroxene and olivine in equilibrium with CO₂-saturated dolomitic carbonate melt at 3 GPa, 1375 °C. The experiments in Di-Ab were designed to bracket those conditions (3 GPa, 1640 °C and 0.8 GPa, 1375 °C), and so minimise the contribution of differential temperature and pressure to partitioning. Partition coefficients, determined by SIMS analysis of run products, differ markedly for some elements between Di-Ab and Di-Ab-Dmt systems. Notably, in the carbonate system clinopyroxene-melt partition coefficients for Si, Al, Ga, heavy REE, Ti and Zr are higher by factors of 5 to 200 than in the silicate system. Conversely, partition coefficients for Nb, light REE, alkali metals and alkaline earths show much less fractionation (<3). The observed differences compare quantitatively with experimental data on partitioning between immiscible carbonate and silicate melts, indicating that changes in melt chemistry provide the dominant control on variation in partition coefficients in this case. The importance of melt chemistry in controlling several aspects of element partitioning is discussed in light of the energetics of the partitioning process. The compositions of clinopyroxene and carbonate melt in our experiments closely match those of near-solidus melts and crystals in CMAS-CO₂ at 3 GPa, suggesting that our partition coefficients have direct relevance to melting of carbonated mantle lherzolite. Melts so produced will be characterised by elevated incompatible trace element concentrations, due to the low degrees of melting involved, but marked depletions of Ti and Zr, and fractionated REE patterns. These are common features of natural carbonatites. The different behaviour of trace elements in carbonate and silicate systems will lead to contrasted styles of trace element metasomatism in the mantle. Received: 15 July 1999 / Accepted: 18 February 2000     

The importance of melt extraction for tracing mantle heterogeneity

Numerous isotope and trace element studies of mantle rocks and oceanic basalts show that the Earth’s mantle is heterogeneous. The isotopic variability in oceanic basalts indicates that most mantle sources consist of complex assemblages of two or more components with isolated long-term chemical evolution, on both global and local scales. The range in isotope and highly incompatible element ratios observed in oceanic basalts is commonly assumed to directly reflect that of their mantle sources. Accordingly, the end-points of isotope arrays are taken to represent the isotopic composition of the different components in the underlying mantle, which is then used to deduce the origin of mantle heterogeneity. Here, a melting model for heterogeneous mantle sources is presented that investigates how and to what extent isotope and trace element signatures are conveyed from source to melt. We model melting of a pyroxenite-bearing peridotite using recent experimental constrains for melting and partitioning of pyroxenite and peridotite. Identification of specific pyroxenite melting signatures allows finger-printing of pyroxenite melts and confirm the importance of lithological heterogeneity in the Earth’s mantle. The model results and the comparison of the calculated and observed trace element-isotope systematics in selected MORB and OIB suites (e.g. from the East Pacific Rise, Iceland, Tristan da Cunha, Gough and St.Helena) further show that factors such as the relative abundance of different source components, their difference in solidus temperature, and especially the extent, style and depth range of melt aggregation fundamentally influence the relationship between key trace element and isotope ratios (e.g. Ba/Th, La/Nb, Sr/Nd, La/Sm, Sm/Yb, ¹⁴³Nd/¹⁴⁴Nd). The reason for this is that any heterogeneity present in the mantle is averaged or, depending on the effectiveness of the melt mixing process, even homogenized during melting and melt extraction. Hence to what degree mantle heterogeneity is reflected in the erupted melts is not only a function of source and melting-induced variability. It also depends on the extent of mixing during melting and melt extraction and thus strongly on the relative incompatibility of the elements considered. The observed trace element variation in erupted melts can be greater or smaller than that of their mantle sources, depending on the incompatibility of the elements investigated. The isotopic variability in erupted melts, on the other hand, is generally smaller than that of their mantle source. Melt mixing during melt extraction consequently has an important influence on the relative extent of variation, and hence the degree of correlation between the isotope and trace element ratios. Overall fewer correlations between trace element and isotope ratios are expected whenever melts are extracted from a restricted depth range, as is the case for ocean island basalts, than for cases where melts are extracted over a larger depth interval (mid ocean ridges and especially ridge centered hotspots like Iceland). While the isotopic composition of the most enriched melts may correspond closely to those of the enriched source component, even the most depleted mid ocean ridge basalts are likely to underestimate the isotopic depletion of the depleted mantle component. These observations imply that using the chemical and isotopic range observed in oceanic basalts as directly representative of that in the corresponding mantle source can be misleading, since this assumption is strictly true only for homogeneous mantle sources. In addition to identifying source or partitioning-related differences in melts from different mantle sources, inferring the true composition, origin, and distribution of heterogeneous components in the Earth’s mantle therefore requires detailed knowledge about the mechanisms of melting and melt mixing during the melt extraction process. Only if these processes and their influence on the isotope-trace element relationship are understood, can the composition and origin of the different source components, and thus mantle heterogeneity, be accurately constrained.     

CH4 inclusions in orogenic harzburgite: Evidence for reduced slab fluids and implication for redox melting in mantle wedge

Fluids released from the subducting oceanic lithosphere are generally accepted to cause mantle wedge peridotite melting that produces arc magmas. These fluids have long been considered to be dominated by highly oxidized H₂O and CO₂ as inferred from erupted arc lavas. This inference is also consistent with the geochemistry of peridotite xenoliths in some arc basalts. However, the exact nature of these fluids in the mantle wedge melting region is unknown. Here, we report observations of abundant CH₄ + C + H₂ fluid inclusions in olivine of a fresh orogenic harzburgite in the Early Paleozoic Qilian suture zone in Northwest China. The petrotectonic association suggests that this harzburgite body represents a remnant of a Paleozoic mantle wedge exhumed subsequently in response to the tectonic collision. The mineralogy, mineral compositions and bulk-rock trace element systematics of the harzburgite corroborate further that the harzburgite represents a high-degree melting residue in a mantle wedge environment. Furthermore, existing and new C, He, Ne and Ar isotopes of these fluid inclusions are consistent with their being of shallow (i.e., crustal vs. deep mantle) origin, likely released from serpentinized peridotites and sediments of the subducting oceanic lithosphere. These observations, if common to subduction systems, provide additional perspectives on mantle wedge melting and subduction-zone magmatism. That is, mantle wedge melting may in some cases be triggered by redox reactions; the highly reduced (∼ΔFMQ-5, i.e., 5 log units below the fayalite-magnetite-quartz oxygen fugacity buffer) CH₄-rich fluids released from the subducting slab interact with the relatively oxidized (∼ΔFMQ-1) mantle wedge peridotite, producing H₂O and CO₂ that then lowers the solidus and incites partial melting for arc magmatism. The significance of slab-component contribution to the geochemistry of arc magmatism would depend on elemental selection and solubility in highly reduced fluids, for which experimental data are needed. We do not advocate the above to be the primary mechanism of arc magmatism, but we do suggest that the observed highly reduced fluids are present in mantle wedge peridotites and their potential roles in arc magmatism need attention.     

Chemical and mineralogical evidence of the occurrence of mantle metasomatism by carbonate-rich melts in an oceanic environment (Santiago Island, Cape Verde)

Lavas from Santiago Island attest to a complex magmatic history, in which heterogeneous mantle source(s) and the interactions of advecting magmas with thick metasomatised oceanic lithosphere played an important role in the observed isotopic and trace element signatures. Young (<3.3 Ma) primitive lavas from Santiago Island are characterised by pronounced negative K anomalies and trace element systematics indicating that during partial melting D_K>D_Ce. These features suggest equilibration with an oceanic lithospheric mantle containing K-rich hydrous mineral assemblages, consistent with the occurrence of amphibole + phlogopite in associated metasomatised lherzolite xenoliths, where orthopyroxene is partially replaced by newly formed olivine + (CO₂ + spinel + carbonate inclusion-rich) clinopyroxene. Metasomatism induced a decrease in $ a ^{{{\text{melt}}}}_{{{\text{SiO}}_{{\text{2}}} }} $ and Ti/Eu ratios, as well as an increase in fO ₂ , Ca/Sc and Sr/Sm in the Santiago magmas, suggesting a carbonatitic composition for the metasomatic agent. Santiago primitive lavas are highly enriched in incompatible elements and show a moderate range in isotopic compositions (⁸⁷Sr/⁸⁶Sr?=?0.70318–0.70391, ¹⁴³Nd/¹⁴⁴Nd?=?0.51261–0.51287, ¹⁷⁶Hf/¹⁷⁷Hf?=?0.28284–0.28297). Elemental and isotopic signatures suggest the involvement of HIMU and EM1-type mantle end-members, in agreement with the overall isotopic characteristics of the southern Cape Verde Islands. The overall geochemical characteristics of lavas from Santiago Island allow us to consider the EM1-like end-member as resulting from the involvement of subcontinental lithospheric mantle in the genesis of magmas on Santiago.