David S. Draper

CONSTRAINTS ON THE ORIGIN OF THE OXIDATION STATE OF MANTLE OVERLYING SUBDUCTION ZONES : AN EXAMPLE FROM SIMCOE, WASHINGTON, USA

Type I spinel peridotite xenoliths from Simcoe Volcano, southern Washington (USA), are from lithospheric mantle approximately 65 km inboard from the axis of the subduction-related Cascade Range. Oxygen fugacities calculated from contents of Fe3+/ΣFe in Simcoe spinels, determined by Mossbauer spectroscopy, are up to 1.4 log units more oxidizing than the FMQ buffer. These are among the most oxidized mantle xenoliths reported, with fugacities substantially higher than those calculated for mantle beneath most of western North America. These results, together with those from amphibole-bearing spinel peridotites from Ichinomegata, Japan (Wood and Virgo, 1989), provide evidence that the mantle above subduction zones is more oxidized than is oceanic or ancient cratonic mantle. We suggest that oxidation was accomplished by an agent ranging in composition from solute-rich hydrous fluid to water-bearing silicate melt. A qualitative model relating extent of oxidation, duration of the oxidation process, and proportion of the available water (derived from subducting slabs) that oxidizes Fe in subarc mantle peridotite, suggests that such an agent can easily produce the observed extents of oxidation over timescales similar to the typical lifespans of subduction zones. For the Cascade arc with a duration of 50 Ma, the observed oxidation in the Simcoe peridotites can be achieved by reacting about 6–11 % of the available water with the mantle. These results demonstrate that water can make an efficient oxidizing agent, and because of the comparatively low ferric iron contents reported for mantle peridotites from other tectonic settings, oxidation of the mantle by water is mostly restricted to subduction zones where water is recycled from the surface and transferred into the mantle wedge.

Experimental insights on crystal chemistry of high-Ti garnets from garnet-melt partitioning of rare-earth and high-field-strength elements

Abstract High-temperature experiments were performed from 3.0 to 7.0 GPa to investigate the effect of composition on near-liquidus garnet-melt trace-element partition coefficients (D-values) in a Ti- and Fe-rich lunar bulk composition. Starting compositions were doped with Sc, Sr, Y, Zr, Ba, Nd, Sm, Dy, Yb, Hf, and Th. D-values were measured by ion-microprobe analysis. The lattice strain model of Blundy and Wood (1994) was applied to measured D-values, and then compared with the predictive garnet-melt trace-element partitioning model of van Westrenen et al. (2001b). Although the latticestrain model describes our data adequately, there is a substantial mismatch between the prediction of the van Westrenen et al. (2001b) model and the actual measured D-values. We suggest crystal-chemical effects associated with the high Ti content of our experimental garnets may be responsible for this mismatch. Titanium alone does not appear to control partitioning; we infer that Fe2+ and/or Mg2+ are required to partition into octahedral coordination on the Y-site, which may further necessitate Si2+ to also partition on the Y-site for charge balance. This combination of effects influences trace-element partitioning in the garnet. Observed changes in D-values correlate with the integration of Ti on the Y-site in a majorite-like exchange, yet our garnets incorporate little or no majorite. Garnets grown have some of the lowest apparent Young.s modulus values (274.528 GPa) yet documented for the garnet X-site indicating higher compressiblities than previously inferred.

Silicic lunar volcanism: Testing the crustal melting model

Abstract Lunar silicic rocks were first identified by granitic fragments found in samples brought to Earth by the Apollo missions, followed by the discovery of silicic domes on the lunar surface through remote sensing. Although these silicic lithologies are thought to make up a small portion of the lunar crust, their presence indicates that lunar crustal evolution is more complex than originally thought. Models currently used to describe the formation of silicic lithologies on the Moon include in situ differentiation of a magma, magma differentiation with silicate liquid immiscibility, and partial melting of the crust. This study focuses on testing a crustal melting model through partial melting experiments on compositions representing lithologies spatially associated with the silicic domes. The experiments were guided by the results of modeling melting temperatures and residual melt compositions of possible protoliths for lunar silicic rocks using the thermodynamic modeling software, rhyolite-MELTS. Rhyolite-MELTS simulations predict liquidus temperatures of 950–1040 °C for lunar granites under anhydrous conditions, which guided the temperature range for the experiments. Monzogabbro, alkali gabbronorite, and KREEP basalt were identified as potential protoliths due to their ages, locations on the Moon (i.e., located near observed silicic domes), chemically evolved compositions, and the results from rhyolite-MELTS modeling. Partial melting experiments, using mixtures of reagent grade oxide powders representing bulk rock compositions of these rock types, were carried out at atmospheric pressure over the temperature range of 900–1100 °C. Because all lunar granite samples and remotely sensed domes have an elevated abundance of Th, some of the mixtures were doped with Th to observe its partitioning behavior. Run products show that at temperatures of 1050 and 1100 °C, melts of the three protoliths are not silicic in nature (i.e., they have <63 wt% SiO2). By 1000 °C, melts of both monzogabbro and alkali gabbronorite approach the composition of granite, but are also characterized by immiscible Si-rich and Fe-rich liquids. Furthermore, Th strongly partitions into the Fe-rich, and not the Si-rich glass in all experimental runs. Our work provides important constraints on the mechanism of silicic melt formation on the Moon. The observed high-Th content of lunar granite is difficult to explain by silicate liquid immiscibility, because through this process, Th is not fractionated into the Si-rich phase. Results of our experiments and modeling suggests that silicic lunar rocks could be produced from monzogabbro and alkali gabbronorite protoliths by partial melting at T < 1000 °C. Additionally, we speculate that at higher pressures (P ≥ 0.005 GPa), the observed immiscibility in the partial melting experiments would be suppressed.