No abstract

The low concentrations of niobium, tantalum and titanium observed in island-arc basalts are thought to result from modification of the sub-arc mantle by a metasomatic agent, deficient in these elements, that originates from within the subducted oceanic crust. Whether this agent is an hydrous fluid or a silica-rich melt has been discussed using mainly a trace-element approach and related to variable thermal regimes of subduction zones. Melting of basalt in the absence of fluid both requires high temperatures and yields melt compositions unlike those found in most modern or Mesozoic island arcs. Thus, metasomatism by fluids has been thought to be the most common situation. Here, however, we show that the melting of basalt under both H2O-added and low-temperature conditions can yield extremely alkali-rich silicic liquids, the alkali content of which increases with pressure. These liquids are deficient in titanium and in the elements niobium and tantalum and are virtually identical to glasses preserved in mantle xenoliths found in subduction zones and to veins found in exhumed metamorphic terranes of fossil convergent zones. We also found that the interaction between such liquids and mantle olivine produces modal mineralogies that are identical to those observed in metasomatized Alpine-type peridotites. We therefore suggest that mantle metasomatism by slab-derived melt is a more common process than previously thought.

Crystallization (dacite) and interaction (dacite-peridotite) experiments have been performed on the 1991 Pinatubo dacite (Luzon Island, Philippines) to constrain its petrogenesis. In the dacite-H 2 O system at 960 MPa, magnetite and either clinopyroxene (low H 2 O) or amphibole (high H 2 O) are the liquidus phases. No garnet is observed at this pressure. Dacite-peridotite interaction at 920 MPa produces massive orthopyroxene crystallization, in addition to amphibole ± phlogopite. Amphibole crystallizing in dacite at 960 MPa has the same composition as the aluminium-rich hornblende preserved in the cores of amphibole phenocrysts in the 1991 dacite, suggesting a high-pressure stage of dacite crystallization with high melt H 2 O contents (>10 wt %) at relatively low temperature (<950°C). The compositions of plagioclase, amphibole and melt inclusion suggest that the Pinatubo dacite was water-rich, oxidized and not much hotter than 900°C, when emplaced into the shallow magma reservoir in which most phenocrysts precipitated before the onset of the 1991 eruption. The LREEenriched REE pattern of the whole-rock dacite demands garnet somewhere during its petrogenesis, which in turn suggests high-pressure derivation. Partial melting of subducted oceanic crust yields melts unlike the Pinatubo dacite. Interaction of these slab melts with subarc peridotite is unable to produce a Pinatubo type of dacite, nor is a direct mantle origin conceivable on the basis of our peridotite-dacite interaction experimental results. Dehydration melting of underplated basalts requires unrealistically high temperatures and does not yield dacite with the low FeO/MgO, and high H 2 O, Ni and Cr contents typical of the Pinatubo dacite. The most plausible origin of the Pinatubo dacite is via high-pressure fractionation of a hydrous, oxidized, primitive basalt that crystallized amphibole and garnet upon cooling. Dacite melts produced in this way were directly expelled from the uppermost mantle or lower crust to shallow-level reservoirs from which they erupted occasionally. Magmas such as the Pinatubo dacite may provide evidence for the existence of particularly H 2 O-rich conditions in the sub-arc mantle wedge rather than the melting of the young, hot subducting oceanic plate.