Geochemical Reservoirs and Whole-Mantle Convection

Abstract

There are currently four rather separate approaches to study the evolution and present state of mantle dynamics, (1) geochemical characterization of mantle reservoirs by studying isotopes and trace elements in mantle rocks and mantle-derived melts, (2) high-pressure mineralogy, (3) numerical simulation of mantle convection, and (4) seismic tomography. Relevant geochemical data have been accumulating over the past 30 years, but quantity, quality and isotopic variety of geochemical data have increased dramatically within the past ten years. Similarly, increasingly rapid progress has been made in the fields of high pressure mineralogy and convection studies, but the most dramatic advance, at least from the point of view of an 'outsider', has come from seismic tomography during the past four years (Grand, 1994; van der Hilst et al. , 1997). The detailed delineation, by independent methods, of high-velocity 'slabs' reaching through the 660 km seismic discontinuity and into the deep mantle is widely accepted as prima-facie evidence for deep subduction and, by implication, for whole-mantle convection. High-pressure mineralogy has shown that the 660 km discontinuity corresponds to a major mineralogical phase transition from ringwoodite (= y-phase olivine) to a Mg-perovskitemagnesiowiistite assemblage in the deep mantle. Most workers in that field now agree that no difference in bulk major-element composition between upper and lower mantle is required to satisfy mineralogical and geophysical constraints. However, whole-mantle circulation could still be inhibited by a dramatic shift in viscosity or by a strongly negative slope ('Clapeyron slope') of the ringwoodite-perovskite transition. Numerical modelers of mantle convection, perhaps partly as a result of the seismic evidence, are also increasingly concerned with whole mantle circulation, and they have been developing models whereby the 660 km seismic discontinuity may retard or temporarily stall vertical movement, rather than inhibiting circulation between upper and lower mantle altogether. To be sure, there are holdouts and heretics in the fields of geophysics and mineralogy who stick to (compositionally and convectively) strongly layered mantle models, but the majority of workers in these fields are now worshipping at the altar of whole-mantle convection. Geochemists, in contrast, have been spoiling the fun of uninhibited whole-mantle convection by pointing to the evidence from noble gas and other (no less noble) isotopes, which require the long-term persistence of separate geochemical reservoirs, and are most easily explained if the upper and lower mantle layers convect separately. This isolation may be disturbed by leaks through the 660 km boundary, but these leaks must be small enough to keep the upper and lower mantle reservoirs from being homogenized. In short, the lnew paradigm' explaining how Earth really works still eludes us, as Dev. L. Advocate (1998) has reminded us. What does geochemical evidence tell us? To start with, geochemistry p e r se provides no evidence for layering. Only in conjunction with other observations and their interpretation can geochemical evidence be interpreted in terms of Earth models. For example, we do not know from geochemistry that the source reservoir of mid-ocean ridge basalts is located in the upper mantle. Rather, we infer this from the observation that the lateral movements of spreading ridges appear to be controlled by plate geometry and movements, in other words by near-surface phenomena. Ridges respond passively to these controls, and common sense seems to dictate that they tap the asthenosphere lying directly beneath the plates. In contrast, hotspots or plumes, whether they are stationary of not, are strongly decoupled from plate movements. The common-sense inference is therefore that they are derived from deeper levels. In any case, it should be remembered that none of this is dictated by geochemistry. What geochemical evidence does tell us is that (1) sources of hot spot magmatism are different from those of mid-ocean ridge basalts; (2) the source reservoirs have in most cases been kept separate for several hundred million years, in some cases for as much as 1.5 to 2 Ga (Austral-Cook and St. Helena hotspot sources); (3) not all of the mantle is as depleted in incompatible trace elements as the MORB source, and the remainder is not made up entirely of hot-spot