DI43A-01 INVITED
CO2 and potassium in the mantle: carbonaceous pelite melts from the trailing edge of a detached slab hybridizing in the mantle to ultrapotassic kamafugite
The ultrapotassic magmas from the Intra-Apennine and Roman provinces constitute worldwide endmembers in terms of K2O/Na2O, K2O content and CO2 degassing, and are associated with carbonatites. Group II kimberlites, which are geochemically similar but less extreme, occur on cratons stable since several 100 Ma. This geotectonic situation appears in strong contrast to the subduction setting of central Italy, where plate convergence has slowed down to less than a few mm/a, the slab now tearing off leading hot asthenospheric mantle to flow in between the trailing slab and the crust. A successful recipe for ultrapotassic magmas requires K/Na fractionation at some previous stage. Melts from fluid-absent melting of carbonaceous pelites at >3 GPa are ultrapottasic phonolites (SiO2 ~64 wt%) and have K2O/Na2O up to 9 because of residual cpx with jadeite80. The effect of CO2 is to stabilize residual jadeite, to lower SiO2, and to increase K2O/Na2O ratios (as compared to CO2 free melts with K2O/Na2O of 1-3 and SiO2 = 73-77 wt%). The carbonaceous pelite melts were equilibrated with fertile, refractory but cpx bearing mantle, and wherlite. At sufficient pressures (3.5 GPa) and XCO2 in the volatile component, hybridization of the carbonaceous pelite melts produces highly subsilicic kamafugites, with K2O/Na2O only slightly lowered, and XMg's >0.70 as characteristic of primitive melts. The essential role of CO2 is to reduce the olivine saturation volume and to shift the olivine-cpx-opx cotectic to lower SiO2. The Italian kamafugites are ultracalcic (CaO/Al2O3 = 1.2-1.4), and although carbonaceous pelite melts have little CaO and 20 wt% Al2O3, the assimilation of cpx and production of aluminous opx leads to ultracalcic compositions when equilibrated with refractory peridotite or wherlite. Temperatures necessary for the fluid absent carbonaceous pelite melting are 1050-1150 °C, far above any reasonable subduction geotherm. Hybridization in the mantle requires 1320-1400 °C (at 3.0-3.5 GPa). The proposition is, that the carbonaceous pelite melts only form in thermally relaxing slabs typical for an ending subduction. Slab break off causes inflow of hot asthenospheric mantle, which further heats the trailing edge of the slab. Normally, subduction ends through continental collision at the surface, and the melts from the trailing edge appear to freeze in the relatively cool mantle below the orogen and await the next thermal event in order to form group II kimberlites. In the particular Italian situation, where large scale plate tectonic movements changed the convergence into a strike slip, temperatures of the trailing slab edge and the asthenosphere are apparently sufficient to keep the carbonaceaous slab melts liquid during hydridization in the mantle, leading to the surface kamafugites. The short longevity of the ultrapotassic magmatism (800 ka) collaborates the concept of a one-time event, i.e. the one-time melting of the trailing slab edge during it's break-off.
DI43A-02
Volatiles and Trace Elements in Rejuvenated-Stage Lavas from Ni'ihau, Hawai'i: Evidence for Silicate Melt and Carbonatite Metasomatism of the Mantle Surrounding the Hawai'ian Plume
We present new volatile, trace element, and radiogenic isotopic compositions for submarine rejuvenated-stage lavas erupted on the northwest flank of Ni'ihau. These lavas are mildly alkalic and are isotopically similar to rejuvenated-stage lavas from Ni'ihau and other Hawai'ian islands. They display trace element heterogeneity, however, greater than that of rejuvenated-stage lavas from other islands, with enrichments in Ba, Sr, and light- rare earth elements resulting in high and highly variable Ba/Th and Sr/Nd. The highest Ba/Th (354) are in the least silica-undersaturated lavas, implying that the greatest enrichments are associated with the largest extents of melting. Ni'ihau lavas also have high and variable H2O/Ce and Cl/La (620 and 39, respectively). We model the trace element concentrations of most rejuvenated-stage lavas by small degrees (1-15%) of melting of depleted peridotite recently metasomatized by a few percent of an enriched incipient melt (0.5-1% melting) from the Hawai'ian plume, similar to the model of Yang et al. (2003, J.Pet. 44). Compositions of Ni'ihau rejuvenated- stage lavas are best explained by addition of 0.1 to 0.4% carbonatite, similar in composition of oceanic carbonatites from the Canary and Cape Verde Islands, to a depleted peridotite less enriched in incipient melt than that for other rejuvenated-stage lavas. Primary H2O and Cl of carbonatite component must be high, but variability in the volatile data may be caused by heterogeneity in the carbonatite composition and/or interaction with seawater. Our model is consistent with predictions based on carbonated eclogite and peridotite melting experiments (Dasgupta et al., 2006, J.Pet. 47) in which 1) carbonated eclogite within the Hawai'ian plume is the first to melt during plume ascent, 2) carbonatite melt metasomatizes the plume itself and the surrounding asthenosphere, 3) as the plume rises, silica-undersaturated silicate melts are also produced and contribute to the metasomatic signature. The metasomatic component is only preserved at the margins of the plume, where low extents of melting of the metasomatized depleted mantle surrounding the plume are sampled during flexural uplift.
DI43A-03
CO2-Brine Mantle Metasomatic Fluids in the Lithosphere Beneath the Ethiopian Plateau
Fluid inclusions in amphibole-bearing spinel lherzolite xenoliths from Quaternary volcanic necks at Injibara (western part of the Ethiopian plateau) show compelling evidence for brines in the lithospheric mantle in a region of continental flood basalts. The petrography of the xenoliths indicates a transition from deformed (protogranular to porphyroclastic texture) Cl-rich pargasite metasomatized lherzolites (LREE, LILE, and Pb enrichment in clinopyroxene) at 950 - 1015°C, to cryptically metasomatized equigranular lherzolites (LREE flat to depleted in clinopyroxene), at higher temperatures (1050 - 1150°C). All pargasite-bearing lherzolites contain CO2-brine fluid inclusions, but no melt inclusions. Fluid inclusions are preserved only in orthopyroxene, while in clinopyroxene and olivine, inclusions underwent extensive post- entrapment interaction with host mineral. The fluid inclusion composition is: XCO2 = 0.64, XH2O = 0.33, XNa = 0.006, XMg = 0.006, XCl = 0.018, (salinity = 14 - 10 NaCl eq. wt. %), although water contents in the originally trapped fluids could have been higher. Trapping pressures are of about 1.4 GPa (at T = 1000°C). Synchrotron sourced micro-infrared mapping (ELECTRA, Trieste, Italy) in nominally anhydrous minerals shows gradients for H distribution, with considerable enrichment at grain boundaries, and along intragranular microfractures. The petrography, major and trace element data, together with constraints from fluid inclusions and microinfrared analyses, indicate two consecutive metasomatic events: a first event of interaction between lherzolite and a C-O-H fluid phase which is Cl-rich, followed by lherzolite interaction with high fractions of basaltic melts, during heating. Present work points towards solute-rich aqueous fluids being important metasomatic agents beneath the Ethiopian plateau, locally forming hydrous regions, which may be easily melted. Presence of high-chlorine in fluids suggests a contribution of recycled altered oceanic lithosphere in their source.
DI43A-04
Melting of Hydrous, Carbonate-bearing Mantle Peridotite
At least some carbonatites contain significant amounts of H2O. This could have important implications for phase relations at the generation of primary carbonatite melts and greatly reduce the solidus temperature of the source. The trace element systematics of carbonatites and the metasomatism associated with their emplacement in the lithosphere could also be affected. There are strong indications that primary carbonatitic melts are produced by very small amount of melting in the asthenosphere, which is experimentally most easily studied in simplified systems. Dalton and Presnall (1998, Contrib. Mineral. Petrol. 131, 123-135) investigated the problem of carbonatite generation by determining the isobarically invariant solidus of carbonate-bearing (dolomite (dmt) at pressures up to 4.8 GPa and magnesite (mst) at higher pressures) garnet lherzolite in the system CaO-MgO-Al2O3-SiO2-CO2 (CMAS-CO2) at 3-7 GPa. In this pressure range, the melts are magnesiocarbonatitic with CaO/(CaO+MgO) of 0.67 and 0.53 at 3 and 7 GPa, respectively. We have initiated an experimental study in this same pressure range to determine the phase equilibria involving the same phase assemblage in the system CMAS-CO2-H2O. In these experiments, the phase equilibria are isobarically univariant. The resulting surface in P-T space, along which CO2- and H2O-bearing melts coexist with the crystalline phase assemblage, fo+en+di+gt+mst/dmt, is bounded at the high-temperature side by the carbonate-bearing garnet lherzolite solidus in the CMAS-CO2 system. As temperatures become lower along the surface, the amount of H2O in the melt steadily increases. So far, we have not reached the low- temperature limits of the surface, and with H2O content of the melts up to >10 wt%, saturation of a water-rich phase has still not occurred nor is liquid immiscibility seen. The quenched melts produced in the experiments are composed of an intergrowth of carbonate and silicate. Because all indications are that H2O exsolves to form a fluid phase during quenching, we have to estimate the H2O content in the melt on the assumption that all the H2O in the starting composition is retained in the sample capsule during experiments and that the amount of H2O incorporated in the crystalline phases is minor compared to its amount in the melt phase. In an experiment at 3.5 GPa, 1100°C, which is 170°C lower than the solidus in the CMAS-CO2 system at the same pressure, the melt contains about 10 wt% H2O. The composition of the melt has shifted toward more calcic composition than anhydrous melts at the same pressure, and with CaO/(CaO+MgO) of 0.70 the composition is starting to approach that of a calciocarbonatite (CaO/(CaO+MgO) >0.8). The amount of SiO2 is also very low, only 1-2 wt%. Hence, it is possible that at lower pressures and with even larger amounts of H2O, calciocarbonatites could be produced by melting of carbonated mantle peridotite. The few data available from higher pressures (up to 7 GPa) indicate that the effect of H2O on melt composition and solidus depression is smaller as pressure increases.
DI43A-05
Volatiles in Basalts From Intra-Transform Spreading Centers: Implications for Melt Migration Models
Volatile elements significantly affect partial mantle melting and their abundance and spatial distribution within the source region provides important constraints on models of melt migration and mantle heterogeneity. An advantage of studying basalts from intra-transform spreading centers (ITSCs) is that high eruption pressures in oceanic transform faults inhibit the degassing of most volatiles, sometimes including CO2. ITSCs also provide well-defined locations of melt delivery and crustal formation because there is no high-level, along-axis melt transport from other parts of the spreading system. In addition, magmatism in fracture zones is probably not supported by long-lived magma chambers. Thus lavas from ITSCs experience less mixing and fractionation than axial MORBs and their compositions provide insight into models of melt migration beneath mid-ocean ridges. The Quebrada/Discovery/Gofar (QDG) fracture zone system offsets the fast-spreading East Pacific Rise (3- 5°S) by approximately 400km and is composed of 7 active ITSCs ranging in length from 5 to 70km. Forty- seven dredges of young intra-transform basalts were collected from this area and analyzed for volatiles, major and trace elements. Incompatible trace element ratios (e.g., Th/La) of QDG lavas range from ultra-depleted to fairly enriched compositions. The level of enrichment correlates well with ratios of volatiles to similarly incompatible refractory elements and with indicators of depth of melt segregation (e.g., Sm/Yb). Overall, the chemical variation of these basalts is greater than that previously found in fracture zones and is similar to the compositional range defined by EPR seamounts (5-15°N). Melt migration models at MORs can be tested by comparing predicted model lava compositions at ITSCs with dredged MORBs. A 3D model detailing mantle temperatures and pressures can be derived by assuming plate- driven passive flow and using measured QDG plate geometries and spreading rates. In conjunction with the pHMELTS model of thermodynamic equilibrium (Asimow et al., 2004), it is possible to construct a 3D array of melt volumes and compositions per cubic km. A simple dynamic pressure gradient model was created from solid flow paths to guide melt from its origin to an ITSC. Melt was traced to its point of eruption, summed, and compositionally contrasted with dredged MORBs. In this model, the deepest melts from beneath the small ITSCs are extracted to nearby large ridge segments, predicting a more depleted lava composition in the transforms. The trace element compositions of QDG lavas are well matched by this simple pressure gradient model using both a depleted and enriched dry mantle source. We will further test this model by comparing the volatile and trace element compositions of the dredged MORBs with the predicted model compositions when water is present in source region.
DI43A-06
Chlorine and Potassium Flux Into the Mantle via Subduction of Oceanic Crust: Constraints From Melt Inclusions in HIMU Lavas
Previous studies have estimated the chlorine content of recycled oceanic crust as ranging from <50 ppm [1] up to ~200 ppm [2,3]. One of the main discrepancies in these estimates is the fraction of K assumed to be retained in the subducted slab. In order to better constrain estimates of Cl and K recycling, we have analyzed olivine-hosted melt inclusions in HIMU lavas from the island of Raivavae to determine Cl concentrations as well as Cl/K2O. K/Nb, and Cl/Nb ratios. Raivavae lavas span a wide range of lead isotopic values, with 206Pb/204Pb ranging from ~19.3 to 21.3. Previous isotopic and trace element studies suggest that the lavas derive from a mantle source containing ancient dehydrated oceanic crust. Chlorine and K2O concentrations range from 40-5300 ppm and 0.22-3.5 wt.% respectively. The majority of Cl/K2O ratios range from 0.01-0.12, Cl/Nb ratios range from 5-25, and K/Nb ratios range from 50-350. Cl/K2O, Cl/Nb, and Cl concentrations are positively correlated with 206Pb/204Pb. Within individual inclusions, there is a positive correlation between Nb/Zr and Cl/Nb. No correlation exists between host olivine forsterite content and Cl/K2O ratios or Cl concentrations. In contrast, K/Nb is negatively correlated with 206Pb/204Pb. The correlation between source signatures, e.g., 206Pb/204Pb and Nb/Zr, with Cl concentrations, Cl/K2O, and Cl/Nb, and lack of correlation with host forsterite content suggest that Cl concentrations are not controlled by shallow assimilation processes. This suggests that the elevated Cl concentrations, Cl/K2O, Cl/Nb are a result of chlorine recycling into the deep mantle and not from assimilation of Cl-rich brine as suggested by [1]. To estimate the amount of chlorine and potassium retained in recycled oceanic crust, we have looked at Cl/Nb and K/Nb ratios because Nb should be retained during slab dehydration. Using an average Cl/Nb ratio of ~18 in HIMU lavas and an average N-MORB Nb content of 5.6 ppm, we estimate ~100 ppm Cl in subducted oceanic crust after slab dehydration, intermediate between previous estimates of [1] and [3]. Using an average K/Nb ratio of ~100 and 5.6 ppm Nb, we obtain ~560 ppm K contained in the subducted slab. This estimate is lower than that assumed by [3]. Given an estimated Cl content in altered oceanic crust of ~157-322 ppm [2] and K content of ~4650 ppm [4], we estimate ~30-65% of the Cl and ~12% of the K in altered crust is retained in the slab and eventually returned to the deep mantle. Using the above estimates of ~100 ppm Cl and ~560 ppm K in subducted oceanic crust we can evaluate the net flux of Cl and K into the deep mantle. Fresh oceanic crust has ~20 ppm Cl whereas subducted crust has ~100 ppm Cl. This suggests a net transfer of Cl into the deep mantle through the plate tectonic cycle. Future Cl isotope measurements of HIMU lavas may allow for further constraints on the amount of Cl in HIMU mantle that is derived from the hydrosphere. In contrast, estimated K in fresh oceanic crust (~500 ppm) is essentially the same as K in subducted crust (~ 560 ppm). Therefore unlike Cl, K is currently in a steady state in the mantle with no net transfer into or out of the deep mantle. [1] Lassiter et al., EPSL 202 (2002) 525-540. [2] Philippot et al., EPSL 161 (1998) 33-44. [3] Stroncik and Haase, Geology 32 (2004) 945-948. [4] Staudigel et al., EPSL 130 (1995) 169-185.
DI43A-07
New Micro-XANES Determinations of Fe Speciation as a Proxy for Oxidation State of Global MORB and Arc Magmas
Erupted lavas may reflect their source oxidation state and offer insight into the relationship between tectonic environment, volatiles, and redox. The Fe3+/ΣFe of basaltic glasses provides one probe of oxidation state and can be determined at high spatial (9μm) and high fO2 (±0.25 log units relative to the QFM buffer) resolution using a new μXANES calibration (Fischer et al., this meeting). We present new μXANES determinations of Fe3+/ΣFe for natural basalt glasses from mid-ocean ridges (n=18) and back-arc basins (n=2), and primitive, undegassed (up to 6.4 wt% H2O), olivine-hosted melt inclusions from arc volcanoes (n=18). We used μXANES to analyze a small, global subset of MORB pillow glasses previously analyzed for Fe3+/ΣFe by colorimetry (Christie et al., 1986). Preliminary fO2 derived from the spectra are significantly more oxidized (~QFM ±0.5) than those determined by Christie et al. (QFM -1.2 ±0.44), but are similar to wet-chemical determinations (Bezos and Humler, 2005; QFM –0.41 ±0.43) for global MORB and the JDFD2 standard. μXANES assures passage of the synchrotron X-ray beam through glass alone, distinguishing it from acid digestion techniques that necessarily sample larger aliquots, including microlites, or that may induce changes in the Fe oxidation state. Inclusion/exclusion of microlites, however, does not appear to account for the higher fO2 measured using μXANES, nor does the X-ray beam have any oxidizing effect on the glass. The systematic offset between Fe3+/ΣFe determined by μXANES and colorimetry could reflect analytical bias or sampling bias due to the contrast in analyzed glass volume between these techniques, though Fe3+/ΣFe varied little within single glass chips (σ~0.02 but up to 0.5 log units). In the olivine-hosted glass inclusions, where small spatial sampling is required, these new data show Fe3+/ΣFe ratios consistently higher (>QFM) than in the global MORB samples, and a general increase in Fe3+/ΣFe with increasing dissolved H2O content (Kelley et al., this meeting).
DI43A-08
Reaction between lherzolite and eclogite-derived melts in the upper mantle
During mantle upwelling, pyroxenite-rich regions are likely to start melting at greater depths than peridotite. As a result, we can anticipate the existence of at least two regimes of chemical and mechanical interaction between peridotites and pyroxenites: one in which pyroxenite is partially molten while peridotite is subsolidus, and the other in which both lithologies are partially molten. In this study we explored the nature of such interactions in both regimes by conducting lherzolite\--pyroxenite-derived melt dissolution experiments. All experiments were performed at 1300°C - 1375°C and 2 GPa in a piston cylinder apparatus using the reaction couple method. At lherzolite subsolidus conditions, the reaction involves negligible amounts dissolution of the lherzolite with crystallization of garnet at the rock-melt interface, shifting the melt towards a more qz-normative composition. The lherzolite is chemically unaffected by the reaction, suggesting that dissolution is rate-limited by the slowest diffusing component in the melt. In contrast, when the lherzolite is partially molten we observed large dissolution rates and the formation of an opx-rich harzburgite + melt layer, sandwiched between the original partially molten lherzolite and pyroxenite-derived melt. The composition of the minerals across the capsule shows the existence of chemical gradients that extend beyond the boundaries of the newly formed lithology. The fast dissolution rates and chemical gradients indicate that melt is interconnected, as confirmed by BSE images. Therefore the opx-rich region does not represent an impermeable barrier as suggested in previous studies. The critical difference between the two regimes is the physical state of the surrounding lherzolite: subsolidus or partially molten. The P-T conditions and the composition of the pyroxenite-derived liquid are important additional factors that determine, for example, the mineralogy of the reaction boundary layer, the dissolution rates, and the time-scale of equilibration. Depending on the style of melt-rock reaction taking place in the mantle, the trace element and isotopic contribution of pyroxenite-derived melts will vary dramatically. During the highP/lowT regime, dissolution of lherzolite will be extremely slow and the pyroxenite-derived melt will react very little with the surrounding mantle, keeping most of its original geochemical signature. As melting of the pyroxenite proceeds, the lack of reaction and infiltration into the peridotite will cause the melt fraction to increase, which could lead to local brittle failure. In order for the pyroxenite-derived melt to reach the base of the lithosphere unaffected by the transport process, a combination of fracturing at highP and flow through pre-existing high-porosity conduits is necessary. Alternatively, extensive re-equilibration will take place, as expected by the lowP/highT regime, and the melt composition will inherit a hybrid signature of both the lherzolite and pyroxenite-derived melts. Using two-phase flow theory applied to a 1\-D mantle column, we calculated the chromatographic effect that an enriched pyroxenite-derived melt would undergo by reacting with a porous mantle. The model shows that the melt composition leaving the porous column evolves from an initial DMM-derived composition towards a variable U- shape trace element pattern in a spider diagram, depending on depth. Although the exact melt compositions produced by this melting model depends on the source composition, transport properties, and P-T path, we believe these initial results show some of the essential features of melt-rock reaction in the mantle.