Petrogenesis of Hawaiian tholeiites: 2, aspects of dynamic melt segregation

To Hawaiian magma genesis, dynamic melt segregation offers a potential resolution of conflict arising between trace-element evidence and phase-equilibria evidence, for deep garnet-present melting versus shallow garnet-absent melting. In this study comprehensive dynamic melting models, which incorporate phase-equilibria constraints and variable partition coefficients, have been applied in efforts to simulate decompression melting of a mantle plume. These models specifically endeavour to reproduce Hawaiian REE (rare-earth-element) patterns from a peridotitic upper mantle source with chondritic relative abundances of middle and HREE (heavy REE). If the flow of both melt and solid mantle is vertical through the partially molten source region, and melting proceeds beyond the stability limit of garnet in peridotite, dynamic melting processes are unable to produce the fractionated REE patterns of Hawaiian tholeiites. Instead, three-dimensional dynamic melting modles need to be invoked, in which lateral migration of the melt relative to the residual matrix also takes place. This enables the derivation of small garnet-equilibrated melt fractions from a larger source volume than that supplying more extensive melt fractions from shallower garnet-absent levels of melting (i.e melting shapes with a mean degree of melting smaller than the maximum extent of melting). This can be achieved by either drawing small-degree melt fractions, formed in the presence of garnet at the plume peripheries, toward the plume centre, or by advecting the mantle residue away from the plume centre as it ascends. Fluid dynamic theory supports a plume model incorporating the latter, with melt flow occurring vertically through a matrix flow which is deflected by the lithosphere and diverges away from the plume centre. In this framework, the generation of melting shapes dominated by small-degree garnetpresent melt fractions, requires a decrease in the rate of melting with progressive melting and height along melt-flow paths within the plume. This is consistent with a decrease in vertical velocity of the matrix (and thus decompression melting rate) upwards through the plume and, with diminishing melting rates upon exhaustion of garnet and clinopyroxene as melting progresses. Providing melt segregation occurs by percolation, equilibrium between the segregating melt and residual peridotite matrix may be maintained throughout the plume. In this way, primary melts extracted from the Hawaiian plume have their bulk compositions determined by phase equilibrium with the extensively melted matrix residue (harzburgite) at the plume top and shallowest level of melting (2.0 GPa), and their incompatible-trace-element characteristics determined by smaller-degree melt fractions derived from deeper, garnet-present levels of melting (3.0 GPa). Simple unidimensional models for melt segregation by percolation or via channels are shown to produce incompatible-trace-element abundances and ratios which are similar to those generated by equivalent degrees of batch melting. Moreover, contrary to a common belief held for dynamic melting, the enrichment of more-incompatible elements over less-incompatible elements is not always greater than that produced by an equivalent amount of batch melting.