Testing the Requirement for Considerable Pyroxenite in the Source of OIB

The presence of a subducted basaltic oceanic crust component (as eclogite) in the source of ocean island basalts (OIB) has long been argued for on the basis of heterogeneity in the radiogenic isotopic composition of OIB (1). This argument has been strengthened by the discovery of high Ni contents in primitive magmatic olivines from Hawaii, seemingly incompatible with a peridotitic source, which should produce low-Ni melts (2). However, the apparent requirement of eclogitic contributions to OIB primary magmas has been difficult to reconcile with the observation that eclogitic melts are typically silica-rich, whereas many OIB are silica-undersaturated. Additionally, OIB on thick lithosphere should bear a stronger eclogite signature, as the lower solidus of eclogite compared to peridotite means it should be preferentially sampled beneath thick lithosphere, where peridotite melting is at a minimum. Yet it has long been observed that OIB on thick lithosphere are the most silica-undersaturated (3), making Hawaii an exception, as an OIB on thick lithosphere with surprisingly silica-rich shield-stage lavas. The trace- and major-element composition of these has been explained by invoking a reaction between peridotite and eclogitic melt in an upwelling mantle plume to produce a secondary, Ni-rich pyroxenite (3). This secondary pyroxenite would then melt during further ascent and mix with peridotitic melt to produce the primary magma for Hawaiian shield-stage volcanism. However, this secondary pyroxenitic melt is still silica-rich (52 wt.% for an 80% melt) (4), so it is difficult to see how it might make a significant contribution to other OIB. Yet it is from the example of Hawaii that a model for the amount of oceanic crust in the source of all OIB has been suggested (4). To test this, we have analysed primitive olivines from La Palma (Canary Islands - thick lithosphere) and São Miguel (Azores - intermediate thickness lithosphere) and fit these to a model of garnet peridotite melting and subsequent melt evolution in terms of olivine forsterite number, Ni, Co and Mn. We show that both La Palma and São Miguel are reproduced in Fo-Ni-Mn space by melting of garnet peridotite alone. With the uncertainty inherent in any model based on parameterisations of partitioning behaviour, no resolvable pyroxenite component is apparent in the source of either island from the Fo-Ni-Mn composition of their olivines. In light of this, we have applied our model to the database used in (4). In Fo-Ni-Mn space, Hawaii is unique among OIB as the only example examined that is not reproduced by our model. Variation in Fo-Ni-Mn composition among OIB is qualitatively linked to degree of melting and appearance of pyroxene in the crystallising assemblage. We observe a strong similarity between Hawaii and continental flood basalts (CFB). Noting that Hawaii is underlain by a particularly vigorous plume, and the common belief that CFB are produced by hot, high-flux plume heads, we suggest a mechanism of deep-sourced melts rapidly equilibrating with shallow lithosphere in high-flux plumes to explain the high Ni content of these lavas. (1) Hofmann & White (1982) EPSL 57, 421-436 (2) Sobolev et al. (2005) Nature 434, 590-597 (3) McBirney & Gass (1967) EPSL 2, 265-276 (4) Sobolev et al. (2007) Science 316, 412-417