A MELT EXTRACTION MODEL BASED ON STRUCTURAL STUDIES IN MANTLE PERIDOTITES

Abstract

This study, largely based on field observations in various peridotite massifs and on basalt xenoliths, traces the successive stages of melt extraction from mantle diapirs. The first stage, initiated in the garnet lherzolite field and pursued in the spinel lherzolite field, creates melt pockets at the sites of the former garnets. During ascent, the melt fraction stable in spinel-lherzolites is estimated to be IJ 7 per cent. Above this fraction, but depending upon plastic strain, permeability is obtained and melt extraction starts. This occurs at 50 km depth. A network of connected melt veins and gashes, opened by fluid assisted shear fracturing in the deforming peridotites, is first created. When its vertical extension attains IJ 10 km. the melt pressure fractures the overlying peridotites (tensional hydrofracturing) creating a conduit for melt extraction. The buoyant forces generate a negative pressure at the base of the conduit. After communication with the earth surface is achieved, the melt network surrounding the dyke root is thus drained. This mechanism explains the remarkable efficiency of melt extraction in residual harzburgites and dunites. The conduit is a dyke, with a 20cm width at shallow depth. The melt velocity through such dykes in shallow mantle is 5 cm s?1. The rate of extraction of melt is so large that melt extraction is necessarily a discontinuous process even in the case of oceanic crust generation. Each dyke of the dyke swarm in oceanic crust and ophiolites (and possibly each cumulate layer in the underlying mafic cumulates) corresponds to a melt extraction event. Thus each event creates 1 m of crust, during the time lapse of a few weeks. The periodicity of such events (5–50 yr) depends on the spreading rate (10–1 cm yr?1). Each one corresponds, in the rising diapir, to a hydrofracture produced locally in the area of the mantle melt network. For spreading rates > 1 cm yr?1, a 6 km thick oceanic crust is created with an underlying uppermost mantle composed of residual harzburgites. For smaller rates, the oceanic crust is thinner as a result of smaller degrees of melting, with plagioclase lherzolites as residue. For even smaller rates, no oceanic crust is created (continental rifting) and the residue is a comparatively fertile spinel lherzolite. This is explained by a direct relation between spreading rate and ascent rate of the mantle diapir. For spreading rates 1 cm yr?1, the adiabatic melting in the diapir stops at about 15 km depth in plagioclase lherzolites (except for a final melt extraction just below the Moho) and at > 30km in spinel lherzolites. This model has implications on melting processes (disequilibrium melting), the nature of primary melts and implies a straighter connection than generally accepted between the physics and chemistry of volcanism and melting processes in the mantle.