THE ROLE OF DUNITES FOR EXTRACTION OF MELTS FROM THE MANTLE TO THE OCEANIC CRUST

Abstract

Dunites ranging in size from centimeter to tens of meter are a typical component of the shallow upper mantle as exposed in ophiolites, lherzolite massifs, and abyssal peridotites. Because melts generated at high pressure become undersaturated in orthopyroxene at lower pressure by about 1% orthopyroxene per kilobar of decompression, a depletion of peridotites in orthopyroxene is a logical consequence of melt migration. If, in addition, melt migration is channeled, dunite formation is a possible result. Channeling is, indeed, required during melt migration to explain the observed undersaturation of MORB in orthopyroxene, the highly depleted nature of residual mantle rocks and the excess Th activity ratios in MORB melts. How can channeling be accomplished? One way is by mechanical segregation within a deforming mantle matrix (Holtzman et al., 2003). In this case, melt is driven into melt-rich zones by pressure gradients developing between less viscous, melt-rich regions and the more viscous, meltpoor matrix. However, because melt-enriched lenses sweep through the peridotite matrix during ongoing deformation, the required disequilibrium between melt and matrix cannot be accomplished with this mechanism alone. Alternatively, melts may move within fractures forming during brittle response as a result of overstepping of the extensional strength of peridotite via magma buoyancy. Highly tabular dunites are commonly observed in the mantle, suggesting that fractures play a prominent role in melt migration. However, for short-lived fractures it is currently not understood how suitable overpressures can be generated in a ductile matrix, whether sufficient melt can be drained into a fracture, and how wide dunites can form. Fractures are clearly the preferred mechanism for melt migration within the lithosphere (which can be tens of kilometers thick under slow spreading ridges) but in order to explain decimeterscale or wider tabular dunites, fractures must probably be linked to a porous flow stage pre-or postdating a fracture (Suhr, 1999). The most consistent model for melt migration is thus reactive channeling via a feedback between porosity generation and flow (Kelemen et al, 1995). Here, the timescales of compaction, for dunite generation, and for porous flow scale well with requirements for melt transport under ocean ridges. Also, reactive flow can explain the local occurrence of bands of enstatite dunites or strongly orthopyroxene- depleted peridotites which are difficult to explain in a fracture-related mechanism. Most observed dunites show an enrichment of moderately incompatible elements over the host rock coincident with the dunite-host boundary, suggesting that dunites were able to carry a distinct melt signature through the mantle. Some predictions and observations are, however, not easily reconciled with reactive flow. The largest dunites in ophiolites are thought to be the carriers of the most unreacted, most coalesced melt flow (Braun and Kelemen, 2002). However, the data available to me do not suggested that large dunites (> 5 m) carry a distinct chemical signature representing less reacted melts. In addition, large dunites tend to occur in specific settings which are best interpreted as locations adjacent to lithospheric walls and/or at a late stage in the history of a spreading center, for example, their preferred occurrence at the base of the Oman Ophiolite, at segment boundaries in Oman (cf. Boudier and Godard at this meeting), or a relation to late, refractory melts generated in a spreading center in Newfoundland (Suhr et al., 2003). Also, at 15° N at the Mid-Atlantic Ridge, large, diffuse, enstatite bearing dunites with an infiltrative tendency towards host rocks may have form along a deeply located base of the lithosphere. A convincing internal chemical zoning in large dunites as predicted by Spiegelmann et al. (2003) has also not been observed to my knowledge. This may again suggest that large dunites do not show the organized flow as predicted for a high melt flux channel. If most large dunites are not related to a coalescing melt network, as seems possible as explained above, a new problem emerges in that small dunites may not be able to carry the unreacted flux observed at the ridge because of diffusive equilibration with the peridotite host (Braun and Kelemen, 2002). My preferred model is therefore that reactive channeling may evolve into fractures forming out of dunite channels (as already considered by Kelemen et al., 1997) and that fractures may again turn into porous channels after their closure (Suhr, 1999), perhaps even in a cyclical fashion. Veins and clusters of cpx occasionally observed within dunites may represent the transitional stage from porous flow to vein formation. I would also argue that many dunites carry a reaction signature as seen in a relatively high Cr# in spinel (40-60) and low NiO values in olivine (~0.35 wt. %). Large dunites wider than several meters form relatively late in the history of a spreading center and be linked to lithospheric walls. A modified reactive channeling model like this could explain the common occurrence of highly tabular dunites not obviously related to lithospheric conditions, the lack of a relation between size of dunites and their chemistry, the perplexing clustering of dunites in certain areas, without requiring larger dunites to have a higher porosity than smaller dunites as suggested from size-frequency relations by Braun and Kelemen (2002).