On the intermittency and crystallization mechanisms of sub-seafloor magma chambers

TL;DR: In this paper, the authors show that a 3 km high chamber reaches a Rayleigh number of 3 x and a Nusselt number of about 8000, and that the plated layer should grow until the heat loss rate decreases to that supplied by cumulatedepositing convection.

Abstract: Summary. The conventional view of a magma chamber being an essentially permanent feature of a fast-spreading ridge is not compatible with the physics of either oceanic crustal structure or internal magma convection. Cumulates form a large volumetric fraction of many ophiolites and are deposited at the bottom of magma chambers, so they are not available as an insulating layer between magma and seawater. Extrusives and dykes are cracked and highly permeable to hydrothermal convection, so they also do not offer much thermal resistance to the cooling of magma. Only the layer of ‘plated’ or ‘isotropic’ gabbro, often observed between cumulates and dykes, is limited to conductive heat transport; a 0.5 km thickness implies a chamber lifetime of no more than l0kyr km-’ of magma. The intermittency of the chamber on such a short time-scale requires that the plates move apart to make room much more rapidly than they could at the steady spreading rate. Fluctuating magma pressure can achieve such intermittent movement by stress-change diffusion through an elastic lithosphere overlying a viscous asthenophere. However, the space needed implies stress diffusion almost all around the Earth, and this, in turn, requires that the intermittent spreading be substantially synchronized all along a major segment of ridge. The pressure-dependent slopes of equilibrium temperatures between crystal phases and liquid silicate magmas exceed the adiabatic gradients of the magmas themselves by about 1°C kb-’. If this temperature difference is considered the convective drive, a 3 km high chamber reaches a Rayleigh number of 3 x and a Nusselt number of about 8000. The plated layer should grow until the heat loss rate decreases to that supplied by cumulatedepositing convection, implying a plated layer about 0.5 km thick, in agreement with observations. The boundary-layer/turbulent core structure of convection at high Rayleigh and Reynolds numbers is consistent with the formation of banded cumulates when a new type of fluid circulation is taken into account. If a suspended crystal phase is present in the bulk fluid, a ‘slow-convection’ mode is possible, where flow velocities are restricted by the equilibration rate between crystals and fluid, and the temperature profile is determined by the thermodynamics of crystal-fluid equilibria.