DI42A-01 INVITED
Earth's deep H cycle: H isotope evidence from the Manus Basin for complementary recycled reservoirs
Determining the H isotope composition of Earth's reservoirs is critical for evaluating the origin of water on Earth and the extent to which surface and mantle reservoirs have exchanged water over time. The hydrogen isotope composition of the upper mantle (MORB) is relatively well constrained (δD = -80 ± 10‰1). However, ocean island basalts (OIB) show significantly more variability. In particular, if we consider OIBs proposed to have mantle sources containing recycled subduction-related components, we find both higher (e.g., Samoa2, Salas y Gomez3, Iceland4) and lower values (Koolau5) than MORB. We propose that this difference reflects whether the subduction component represents recycled mantle wedge peridotite or the slab itself. Here we present new H isotope data for Manus Basin glasses and show how complementary H isotope reservoirs could be created. The Manus back-arc basin is a complex region where erupted lavas show superimposed plume, MORB and subduction-related components. He isotope studies of submarine glasses6 show high 3He/4He ratios (up to 15RA) consistent with derivation from a lower mantle plume, thought to originate at the core- mantle boundary. Subsequent work7 found anomalously low δ18O values in the high 3He/4He samples, which could reflect interaction of an ancient recycled slab component with the Manus plume. δD values of Manus glasses with the highest 3He/4He and the lowest δ18O are extremely low (down to δD = -126‰). We argue that the low δD values reflect a recycled slab component in the mantle source, rather than degassing processes (based on CO2 abundances and He-Ar systematics). Based on our studies of hydrogen isotopes in Mariana arc melt inclusions8, we find high δD values, consistent with experimental dehydration-induced fractionations which predict release of a D-enriched fluid from the slab into the mantle wedge. Thus, as dehydration proceeds, the slab will evolve to progressively lower δD values, while the wedge will be fluxed with D-enriched water. We suggest that ocean island basalts (OIB) with recycled slab components should be characterized by low δD, while high δD signatures could be derived from OIB containing recycled mantle wedge peridotite. 1Kyser and O'Neil, GCA (1984) 48 2123-2133 2O'Leary, GCA v. 71, Suppl. 1 A737 3Kingsley et al. (2002) G-cubed 3, U23-U48 4Poreda et al. (1986) EPSL v.78 1-17 5Hauri (2002) Chem. Geol. 183 115-141 6Macpherson et al. (1998) Geology 26 1006-1010 7Macpherson et al. (2000) EPSL 176 171-183 8Shaw et al. (2007), submitted
DI42A-02
Extending the Wet Mantle Solidus: Implications for H2O Transport and Subduction Zone Melting Processes
Recent experimental studies (Grove et al., EPSL 249, 74-89, 2006) determine the solidus for primitive undepleted peridotite at H2O -saturated conditions to 3.2 GPa and reveal the presence of hydrous phases on the solidus above 2 GPa. We present new data from piston cylinder and multi-anvil experiments that extend the peridotite solidus to 5 GPa at H2O -saturated conditions. The H2O -saturated solidus extends from 800°C at 3.2 GPa to 840°C at 5 GPa. Olivine, orthopyroxene, clinopyroxene, garnet ± chlorite ± Ti-clinohumite are stable on the solidus from 3.2 to 5 GPa. The hydrous phase chlorite (12 wt% H2O) is stable on the H2O-saturated solidus from 2 GPa to 3.6 GPa. Above 3.6 GPa, the H2O- saturated solidus and chlorite stability fields diverge. The presence of chlorite on the H2O-saturated solidus has important implications for melting processes at subduction zones and H2O storage in the mantle wedge and subducted lithosphere. Our findings suggest chlorite is stable in the mantle wedge. This provides a new mechanism for transporting hydrous fluids released from the slab in the forearc to the base of the mantle wedge. In addition, chlorite stable within the subducted lithospheric mantle may be an important courier of H2O to melting zones beneath arcs. The location of the breakdown of chlorite and the 10Å phase (both >10 wt% H2O) will control the maximum depth attained by the majority of H2O in the subducted lithosphere. This H2O-out reaction in the slab follows the chlorite-out boundary between 2 - 4.5 GPa and the 10Å-out boundary at pressures greater than 4.5 GPa. For the 10Å phase to transport H2O greater than 4.5 GPa, the mantle lithosphere must remain below 650°C up to 6 GPa, a value colder than those predicted by most thermal models. Therefore, all hydrous minerals with >10 wt% H2O in the subducted lithosphere become unstable at depths of 120 - 150 km in most arcs and elucidate the maximum depth of H2O-saturated melting at subduction zones. Ti-clinohumite (~3 wt% H2O) is stable over a wide P-T range in the subducted lithosphere and may transport small amounts of H2O deep into the mantle.
DI42A-03
Hydrogen Disorder and Elasticity of Phase D at High Pressures
One of the major goals of Earth's sciences is to develop models for the evolution of our planet. This goal is directly linked to our understanding of the dynamics within the Earth's interior. It has long been recognized that small amounts of volatiles such as hydrogen can have a disproportionately large effect on viscosity. This implies less resistance to convection and shorter overturn times. Thus knowing the abundance of hydrogen and its distribution in the mantle has important implications for the evolution of our planet. However, one of the remaining question is provide observables that may aid constraining the presence of volatiles. Phase D, MgSi2O6H2, is of particular interest since it is the only hydrogen bearing phase whose stability field extends into the lower mantle. Thus, phase D is a prime candidate for hydrogen transport into the lower mantle along subduction zones. In order to investigate the elasticity of phase D, we performed static (0 K) first-principle calculations. All calculations were performed with a plane-wave basis-set using GGA-PAW potentials. We find that long H-H bonds are energetically favorable which is likely due to H-H repulsion. The equation of state of the energetically most favorable structures are consistent with previous and experimental and theoretical studies of Phase D. Our preliminary results show that hydrogen bond symmetrization occurs in the pressure range of 40-50 GPa in agreement with previous theoretical studies. The elastic constant tensors of the relevant hydrogen distributions agree to within ~0.5% at least up 70 GPa. We also find that the largest change in elasticity with hydrogen symmetrization is an increase of C33 by ~ 10%. The induced changes of compressional and shear wave speeds are similar, ~ 1% at 40 GPa. The predicted azimuthal and polarization anisotropies of shear waves decrease by only ~ 1% with symmetrization. In contrast, the azimuthal P-wave anisotropy decreases from ~ 18% to ~ 7% with hydrogen bond symmetrization. Thus, changes in azimuthal P-wave anisotropy with depth along subduction zones may allow to constrain the presence of Phase D and the transport of hydrogen into the Earth's lower mantle along subduction zones.
DI42A-04
Mineral Specific IR Molar Absorption Coefficients for Routine Water Determination in Olivine, SiO2 polymorphs and Garnet
Conventionally applied Infrared (IR) calibrations [1, 2] for quantitative water analyses in solids are established on hydrous minerals and glasses with several wt% water. These calibrations are based on a negative correlation between the IR molar absorption coefficient (ε) for water and the mean wavenumber of the corresponding OH pattern. The correlation reflects the dependence of the OH band position on the appropriate O- H...O distances and thereby the magnitude of the dipole momentum which is proportional to the band intensity. However, it has been observed that these calibrations can not be adopted to nominally anhydrous minerals (NAMs) [3].To study the potential dependence of ε on structure and chemistry in NAMs we synthesized olivine and SiO2 polymorphs with specific isolated hydroxyl point defects, e.g. quartz, coesite and stishovite with B3++H+=Si4+ and/or Al3++H+=Si4+ substitutions. Experiments were performed with water in excess in piston cylinder and multi-anvil presses. Single crystal IR spectra demonstrate that we successfully managed to seperate generally complex OH patterns as e.g. observed in natural quartz and synthetic coesite. We quantified sample water contents of both natural samples and our run products by applying proton-proton-scattering [4], confocal microRaman spectroscopy [5] and Secondary Ion mass spectrometry. Resulting water concentrations were used to calculate new mineral specific εs. For olivine with the mean wavenumber of 3517 cm-1 we determined an ε value of 41,000±5,000 lmol-1H2Ocm-2. Quantification of olivine with the mean wavenumber of 3550 cm-1 in contrast resulted in an ε value of 47,000±1,000 lmol-1H2Ocm-2. Taking into account previous studies [6, 7] there is evidence to suggest a linear wavenumber dependent correlation for olivine, where ε increases with decreasing wavenumber. In case of the SiO2 system it turns out that the magnitude of ε within one structure type is independent of the liable OH point defect and therewith the wavenumber of the observed band position. Consequently, one single mean ε of 68,000±5,000 lmol-1H2Ocm-2 could be determined for a suite of quartz samples with varying OH point defects. In contrast, ε varies with the structure itself. For polymorphic coesite we calculated a different ε of 214,000±8,000 lmol-1H2O}cm-2, that is in good agreement with earlier established data [8]. Quantification data of stishovite resulted in an even higher value of ε=867,000±29,000 lmol-1H2Ocm-2, similar to that determined by [9]. First data on natural garnet give an ε value of 40,000±2,000 lmol-1H2Ocm-2, that confirms prior suggested values [10]. Our results demonstrate that not using mineral specific calibrations for quantitative water analyses in NAMs leads to overestimation of sample water concentrations, that are required for modelling the earth's deep water cycle. [1]Paterson, M. S. (1982), Bull. Min., 105, 20-29. [2]Libowitzky, E., Rossman, G. R. (1997), Am. Min., 82, 1111- 1115. [3]Rossman, G. R. (2006), Rev. Mineral., 62, 1-28. [4]Reichart et al. (2004), Science, 306, 1537-1540. [5]Thomas et al. (2006), Am. Min., 91, 467-470. [6]Bell et al. (2003), JGR, 108, (B2), 2105-2113. [7]Koch-Mueller et al. (2006), PCM, 33, 276-287. [8]Koch-Mueller et al. (2001), PCM, 28, 693-705. [9]Pawley et al. (1993), Science, 261, 1024-1026. [10]Maldener et al. (2003), PCM, 30, 337-344.
DI42A-05
Hydrogen solubility in garnet at high pressures
Garnet is the most important secondary mineral whose fraction ranges from ~ 20% in the shallow upper mantle for the pyrolite model to ~ 80% in the transition zone for the piclogite model. Therefore understanding the solubility and dissolution mechanisms of hydrogen in garnet is important for water budget as well as for understanding plastic properties of the mantle. However, there is little consensus on the solubility and dissolution mechanisms of hydrogen in garnet. For example, under deep upper mantle conditions (P>7 GPa), Withers et al. (1998) concluded virtually no hydrogen solubility in pyrope whereas Lu and Keppler (1997) showed small hydrogen solubility (~ 200 ppm wt of H2O). In order to address possible causes of such discrepancy, we have initiated a systematic study on hydrogen solubility under a broad range of pressure and controlled chemical environment. In contrast with the previous study, preliminary results indicate that natural garnet crystal surrounded by powder of olivine, orthopyroxene and clinopyroxene, could host around 1000 ppm wt of H2O at 9 GPa and 1100 °C conditions. Possible interpretation is that the enrichment of hydrogen is due to the increased Mg/Si ratio (activity of MgO) due to the coexistence of garnet with olivine. If hydrogarnet substitution (O4H4) is the dominant mechanism of hydrogen dissolution, then the increase in the activity of MgO will increase the hydrogen solubility. Further experiments are performed (i) to investigate the hydrogen solubility under deep upper mantle conditions (P=6-9 GPa T=1373-1573 K) where the previous results have a major discrepancy, and (ii) to investigate the kinetics and the influence of oxide buffer on hydrogen solubility.
DI42A-06
Solubility of Hydrogen in Olivine as a Function of Pressure and Oxygen Fugacity
The incorporation of hydrogen into olivine is influenced by many of thermodynamic variables (pressure, temperature, oxygen fugacity, etc.) [1, 2, 3]. Given the strong influence that water has on the melting and mechanical behavior of mantle peridotite, it is necessary to determine the solubility of hydrogen in olivine over the range of chemical environments found in the upper mantle. We present results from new high temperature water-saturated hydration experiments to determine the effect of pressure and oxygen fugacity on hydrogen solubility in San Carlos olivine at upper mantle conditions. Our results indicate that at 1 to 2 GPa varying the fugacity of oxygen between the Fe-FeO and Ni-NiO buffers produces significantly smaller change in the concentration of hydrogen in the olivine than has been found at in previous experiments carried out at 300 MPa. Experiments were carried out in a piston-cylinder device at 1 to 2 GPa and 1200 °C using natural San Carlos olivine as a starting material. The fugacity of oxygen was controlled at Fe-FeO, FeO-Fe3O4 and Ni-NiO using solid buffers. Water content in experimental products was measured by secondary ionization mass spectrometry. Variable duration experiments indicate that hydrogen is homogeneously distributed in the olivine at 12 hrs. Our experimental results indicate that pressure strongly effects water content, in agreement with previous studies. For example, as pressure increases from 1 to 2 GPa, the water content of olivine increases from 32 ± 3 to 78 ± 7 ppm at the Ni-NiO buffer, and from 29 ± 3 to 69 ± 3 ppm at the FeO- Fe3O4 buffer. At each pressure the water content is only weakly affected by changing oxygen fugacity conditions, a result that disagrees with a recent report of a 5 times increase in water content between Fe-FeO and Ni-NiO at 2 GPa [3]. Differences between studies may result from variable degrees of defect equilibration or from differences in measurement techniques. However, when compared with the relationship between olivine water content and {\f}O2 determined by [2], data from both this study and [3] suggest that the effect of oxygen fugacity on total water content is significantly weaker at upper mantle pressures than at 300 MPa. [1] Kohlstedt, Keppler, and Rubie (1996) Contrib Mineral Petrol 123:345-357. [2] Bai and Kohlstedt (1993) Phys Chem Minerals 19:460-471. [3] Grant et al. (2007) EPSL, doi:10.1016/j.epsl.2007.06.024
DI42A-07
Hydrous Partial Melting of the Upper Mantle as Judged From Mineral/Melt Partition Coefficients.
It has been known for more than 10 years that nominally anhydrous minerals may incite small amounts of melting in Earth's mantle. In recent years, there has been a great increase in experimental data on the H2O storage capacity of nominally anhydrous minerals. Yet, disagreement among models for the locus of dehydration melting in the upper mantle is growing rather than narrowing. Models span the gamut from requiring small amounts of hydrous melt throughout the upper mantle, to hydrous melting in a global low velocity zone layer at depths of ~80-200 km, to melting only beneath ridges and oceanic islands in a restricted interval a few 10s of km beneath the locus of dry melting. These disagreements persist because direct experimental investigations of the influence of small amounts of H2O on mantle melting are not feasible, and consequently understanding comes from parameterization of indirect experimental constraints. One key constraint on the possible locus of hydrous melting in the upper mantle is the range of feasible concentrations of H2O in near-solidus hydrous melts. For a mantle with a fixed amount of H2O, Cmantle, the maximum H2O concentration of an incipient partial melt is given by Cmantle/Dperid/melt, where Dperid/melt is the equilibrium bulk partition coefficient between the peridotite mineral residue and the partial melt. To address this issue, we conducted experimental determinations of Dmineral/melt for upper mantle minerals (garnet, cpx, opx) from 3 to 5 GPa using piston cylinder and multi-anvil devices. For minerals, concentrations of H2O were determined using low-blank SIMS techniques; for glasses, concentrations were determined by SIMS, FTIR, electron microprobe totals, and confocal Raman spectroscopy. Values for Dgarnet/melt span a wide range, and are apparently controlled by the concentration of minor elements in garnet, notably TiO2. Finally, values for Dpyx/melt depend strongly on the concentration of Al in pyroxene, and in particular on the abundance of tetrahedral Al in pyroxene. Combining experimental constraints on Dmineral/melt with the modal proportions and compositions of minerals near the peridotite solidus, it is possible to estimate the H2O concentration of incipient partial melts. For mantle with 100 ppm H2O, such partial melts have 1.2 wt.% H2O at 3 GPa and 2.3 wt.% at 7 GPa. Such modest concentrations of H2O provide only small stabilization of melt relative to a dry peridotite system, thereby supporting the inference that dehydration partial melting of normal mantle is feasible only near where the dry peridotite solidus is approached, meaning immediately below the locus of dry melting beneath ridges and mantle plumes.
DI42A-08
The Infidelity of Melt Inclusions?
Melt inclusions are routinely used as evidence of magmatic compositions prior to volcanic eruptions. However, it has long been known that kinetic processes can modify melt inclusion compositions during trapping. We investigated the fidelity of melt inclusions as records of magmatic compositions by artificially creating melt inclusions through crystallization of plagioclase and clinopyroxene from a hawaiitic basalt bulk composition at 1.0 GPa, 1150 °C, or 75 °C undercooling. We compared melt inclusion compositions to those of melts 100's of μm away from the crystals and found measurable differences in the compositions. We modeled the concentration profiles of Al, Fe, P, S, and Cl in front of the crystals using classical impurity rejection theory during growth at a constant rate followed by a growth hiatus during which diffusive relaxation occurred. The values of the growth rates and times were constrained by measured crystal sizes and the experimental duration. The diffusion coefficients for the elements investigated were either calculated from transition state theory (Al, Fe) or measured in the same bulk composition as that used for this study (S and Cl from the literature, new measurements for P). An envelope of models bracket the observed compositional profiles of the elements studied and predict enrichments, or depletions, at the crystal-melt interface that are quantitatively similar to those seen in the melt inclusions. The differences between the melt far from the crystals and that at the interface are only 20% (relative) for the major elements, Al, Fe, but are about 50% for S and Cl and can exceed 100% for P. These differences correlate with the relative chemical diffusion coefficients of these ions in the melt. Based upon these experiments and models we advise caution in the use of melt inclusions as indicators of pre-eruptive melt compositions.