Untitled Document

Clarification of the Influence of Water on Mantle Wedge Melting

David H Green and William O Hibberson

Research School of Earth Sciences, Australian National University, Canberra, ACT 0200, Australia

Water is a significant component in primitive island arc magmas and its ubiquitous presence is attributed to release of water from dehydration reactions in subducted oceanic crust and lithosphere.  Water released from the subducted slab is inferred to be transferred as aqueous vapour or water-rich melt into the overlying peridotite of the mantle wedge.  Because of the inverted temperature gradient inferred for the mantle wedge immediately above the subducted slab, access of aqueous vapour or water-rich melt will initiate melting close to the water-saturated peridotite solidus.  The location of a region of water-saturated mantle melting, if it exists, can be predicted if we know the P,T dependence of the water-saturated peridotite solidus and can model the temperature distribution in a particular subduction setting.  A number of experimental studies in the 1970’s defined the P,T conditions for the water-saturated solidus of lherzolite. The solidus decreases rapidly from ~1100°C at atmospheric pressure to 0.5 GPa, 1000°C, and continues to decrease slightly to a minimum of  970°C at 1.5 to 2 GPa.  At higher pressure, the water-saturated solidus increases in temperature (i.e. positive dT/dP), so that it is above 1100°C at 4 GPa.

A new experimental study by Grove et al (2006)* suggests major revision of this earlier work such that these authors’ interpretation of their observations placed the water-saturated lherzolite solidus at ~1000°C at 0.5 GPa but their solidus continues with a negative dT/dP to at least 3.5 GPa, 800°C.  Thus, in comparison with earlier work summarised above, Grove et al (2006) conclude that melting at depths around 100km within the mantle wedge begins at >200°C below the previously determined conditions.

Additional major differences in the new work are the presence of chlorite as a subsolidus and above-solidus phase from 2 GPa to 3.2 GPa and the restriction of amphibole (pargasite) to pressures <2 GPa.,  compared with 3GPa found in earlier work.  These differences in observation and interpretation of peridotite melting have large effects for geophysical, geodynamic and petrological modelling of convergent margin processes and characteristics and there is a need for reconciliation of the new and earlier results.

We have therefore repeated selected experiments of Grove et al (2006) and extended their investigation particularly by varying the water content of the experimental charges from the very high value (14.5 wt%) used by Grove et al (2006).   We have carried out experiments which clarify the roles of aqueous fluid (including solid phases quenched from fluid), water-rich silicate melt and amphibole (pargasite) stability. We reproduce the observations of Grove et al. (2006) with 14.5% H2O but demonstrate that glass observed in the experimental charges below 1000°C is quenched from a water-rich fluid phase, and not from a silicate melt. By varying the water content of the charge we are able to differentiate between fluid-saturated silicate melt and the coexisting water-rich fluid-phase for water contents in the peridotite between 0.073% and 14.5% H2O. Pargasite is stable to 3GPa at 1000°C with low water contents (2.9%, 1.45%, 0.145% and 0.073% H2O) and at this P,T, has a modal abundance of  3% of the peridotite. Pargasite is destabilised at higher water contents of 14.5% and 7.25% as sodium and potassium enter the fluid rather than the silicate minerals. Compositions of clinopyroxene and pargasite correlate with water content. At 2.5GPa coexisting melt and fluid have ~25% and ~80% by wt of  H2O respectively. We demonstrate that the fluid-saturated solidus of the lherzolite model mantle composition is at 1000°C<T<1025°C at 2.5 GPa and 1200°C<T<1225°C at 4 GPa.. The near-solidus melt at 2.5 GPa. is a very silica-undersaturated olivine nephelinite.

We also used olivine single crystal discs and either olivine aggregates or carbon sphere aggregates as melt and fluid traps forming interstitial  films or inclusions within olivine. For several experiments with high water contents, the capsule was pierced under high vacuum at room temperature and the vapour released was analysed by gas chromatography.

The study confirms that early work establishing the fluid-saturated solidus of fertile lherzolite+ H2O (model mantle compositions) at ~1025°C at 2.5 GPa and has a positive δT/δP at higher pressure, is correct. Similarly, pargasite is stable at and below the lherzolite+ H2O solidus up to pressures of 3 GPa. Chlorite is not stable at subsolidus or near-solidus conditions in lherzolite unless high water/rock ratios remove the Na2O+K2O components required for pargasite or phlogopite stability. Our new experiments demonstrate that melting of lherzolite+ H2O at pressures to at least 4 GPa does not show supercritical behaviour but above the fluid-saturated solidus at high pressures has water-rich (~30% H2O) silicate melt and alkalis+silica-rich aqueous fluid (~80% H2O) in sufficiently water-rich compositions. In realistic geological settings, including the mantle wedge above a subducted slab, we should expect lherzolite+H2O to experience either fluid-undersaturated melting ( i.e. the solidus of pargasite lherzolite or phlogopite lherzolite ) or fluid-saturated melting with all fluid components entering the melt at, or very close to, the solidus.

 


* Grove,T.L., N. Chatterjee, S.W. Parman, and E. Medard, (2006) The Influence of H2O on Mantle Wedge Melting. Earth and Planetary Science Letters, v.249, 74-89