Modelling komatiitic melt accumulation and segregation in the transition zone

Abstract

Komatiites are probably produced in very hot mantle upwellings or plumes. Under such conditions, melting will take place deep within the upper mantle or even within the mantle transition zone. Due to its compressibility at such pressures, melt might be denser than olivine, but would remain buoyant with respect to a peridotitic mantle both above and below the olivine–wadsleyite phase boundary because of the presence of its higher temperature and denser garnet. We studied the physics of melting and melt segregation within hot upwelling mantle passing through the transition zone, with particular emphasis on the effect of depth-dependent density contrasts between melt and ambient mantle. Assuming a 1D plume, we solved the two-phase flow equations of the melt-matrix system accounting for matrix compaction and porosity-dependent shear and bulk viscosity. We assumed a constant ascent velocity and melt generation rate. In a first model series, the level of neutral buoyancy zneutr is assumed to lie above the depth of onset of melting, i.e. there exists a region where dense melt may lag behind the solid phases within the rising plume. Depending on two non-dimensional numbers (accumulation number Ac, compaction resistance number Cr) we find four regimes: 1) time-dependent melt accumulation in standing porosity waves that scale with the compaction length. The lowermost of these waves broadens with time until a high melt accumulation zone is formed in steady state. During this transient solitary porosity waves may cross the depth of neutral density and escape. 2) steady-state weak melt accumulation near zneutr, 3) no melt accumulation due to small density contrast or, 4) high matrix viscosity. In regime 4 the high mantle viscosity prevents the opening of pore space necessary to accumulate melt. In a second series, the rising mantle crosses the olivine–wadsleyite phase boundary, which imposes a jump in density contrast between melt and ambient mantle. A sharp melt porosity contrast develops and a large melt porosity accumulates immediately above the phase boundary. Both model series show 1) that not only melt density, but also porosity-dependent matrix viscosity controls the melt ascent or accumulation, 2) that there are parameter ranges and physical conditions which may lead to the accumulation of very large melt porosities (> degree of melting), 3) that in spite of melt being denser than olivine at some depths, in general these melts escape these regions and continue to percolate upward faster than the rising mantle. Melting and melt transport under the conditions predicted by the numerical modelling is able to reproduce the compositions of the main types of komatiite. Thus, the accumulation of large melt fractions, and sequential escape of melt from porosity waves, explains several puzzling features of the geochemical compositions of komatiites.