The effects of deep damp melting on mantle flow and melt generation beneath mid-ocean ridges

We explore the implications of experimental constraints on mantle rheology and the thermodynamics of melting for mantle flow and melt generation benea th a mid-ocean ridge. Numerical models are used to investigate the effects of (1) rheologies affected by dehydration during melting, the presence of m elt? and a transition in creep mechanism to grain boundary sliding, and (2) variations in the rate of melt production. Water in the mantle deepens the peridotite solidus, producing a region of damp melting between approximate ly 70 and 120 km depth. Extraction of water during melting can increase man tle viscosity by as much as two orders of magnitude. At the same time, the presence of melt can decrease the mantle viscosity. A decrease in recrystal lized grain size due to the presence of melt can promote a transition in th e dominant deformation mechanism to grain boundary sliding limited by creep on the easiest slip system for olivine, resulting in an additional order o f magnitude decrease in viscosity. The increase in viscosity associated wit h dehydration significantly inhibits buoyant mantle flow in the dry melting region. However, buoyantly driven flow is predicted in the damp melting re gion if the viscosity in this depth interval is on the order of 10(18) Pa s ; a viscosity this low can be achieved if a transition to grain boundary sl iding occurs after the onset of damp melting. Previous models suggest that crustal thickness variation with spreading rate is a consequence of conduct ive cooling at the top of the melting region. If melt productivity is highe r, so that more melt is produced at greater depth, then the influence of sp reading rate on crustal thickness is reduced even in the absence of buoyant upwelling. Rheology has a significant effect on the size and shape of the melting region and the strain distribution beneath the ridge at a given spr eading rate. For example, buoyant upwelling in the damp melting region loca lizes melt production and induces a region of strong elongation in finite s train ellipses between 50-120 km, offering an explanation for the apparent difference in seismic anisotropy inferred from body and surface waves in th e MELT experiment. The models indicate that the magnitudes of the effects o f buoyant flow in the damp melting region become increasingly prominent at slower spreading ridges. If lattice preferred orientations produced by deep buoyant flow are incorporated into the lithosphere as it thickens by condu ctive cooling, then anisotropy in old lithosphere may be greater for lithos phere created at slower spreading rates. (C) 2000 Published by Elsevier Sci ence B.V. All rights reserved.