For many years geochemists have lived uncomfortably with the hypothesis that the mantle consists of several reservoirs
(e.g. DMM, HiMu, EMI, EMII, PREMA, ...) which have different isotopic histories, yet which for the most part have similar trace element abundances so that mixing of these reservoirs before/after melting produces the observed relatively linear isotopic patterns. Here we pursue the potential implication of recycling models of mantle evolution in which mantle differentiation during the oceanic plate cycle has produced discrete heterogeneities that evolve through time to the 'reservoirs' sampled by OIB, but which all coexist, on the typical 10s of meters to kilometer length scales seen in the apparent mantle exposures of the Ronda Massif. In this framework, the plume melting process acts to distill the EMI and EMII fruits from 'typical mantle' to produce the DMM reservoir of previous lherzolite residues of OIB/MORB melting and the processed HiMu reservoir of the residue of MORB-crust that has passed through the OIB crucible. This distillation is governed by the melting behavior of island arc processed and recycled sediments, OIB, MORB, and residual lherzolite and harzburgite which will be in rough thermal
equilibrium but not in mesoscale chemical equilibrium if their mixing scale is O (~0.01-5 km). (Note that, in this paradigm the non-radiogenic rare gases are the only incompatible elements that have a predominant source from the least melted residues of lherzolite (melts) -> OIB-MORB differentiation.) We present the results of two different approaches we are pursuing to try to quantify and assess this model. The first is to look at variations in chemistry along a given hotspot chain and relate these
variations to the progressive melting of plume asthenosphere beneath the lithosphere - e.g. the isotopic depletion of recent (late-stage) Oahu basalts that erupt behind the Hawaiian hotspot is a consequence of most plums being stripped by melting under Hawaii. The second approach is to produce a quantitative evolution model with ~15,000 discrete cells of chemically heterogeneous 'mantle' material which has inputs based on current differentiation and recycling rates. The mantle is randomly sampled by plumes that partially melt and contribute to an asthenosphere that in turn is sampled by ridge melting. We hope to present either a viable model or a strong failure at the conclusion of this talk.
Both the North American and European plates have a predicted northward component of velocity of 4 mm/yr relative to a fixed Iceland hotspot. Depleted mantle produced by melting of the Iceland plume will move northward at this rate. (The more viscous and more buoyant melt-residue will 'adhere' to the plate above, be pulled apart by spreading, and move northward with the common velocity of the two plates relative to the hotspot.) This will create a barrier extending down into the asthenosphere (of order 100 km) which inhibits northward flow of shallow asthenosphere to feed the Kolbeinsey Ridge. This implies the Kolbeinsey Ridge must be fed by deeper mantle flow contrasting to the Reykjanes Ridge with its large shallow along-axis flow. We will try to explain the geochemical differences of the Reykjanes and Kolbeinsey ridges with this model.