Island arc and mid-ocean ridge volcanism, modelled by diapirism from linear source regions

Abstract

The gravitational instability of buoyant regions of partial melt in the upper mantle has been proposed as the underlying cause of regularly-spaced volcanism at both island arcs and mid-ocean ridges. Motivated by this suggestion, we report a laboratory study of diapirism initiated by the release of buoyant fluid from a linear source into a homogeneous, viscous environment. Fixed-volume and fixed-flux releases of buoyant fluid are studied. A fixed-volume release was obtained by rapidly towing a point source of fluid through a body of denser fluid. The horizontal cylinder of fluid, which was left behind by the source, subsequently underwent a gravitational instability yielding plumes with a characteristic spacing. Expressions for the mean spacing of the plumes and for the growth rate of instability are derived by dimensional arguments, and are compared with the experimental results. Both quantities are found to be independent of the viscosity of the buoyant fluid, a surprising result which we have confirmed theoretically. In contrast, the plume spacing and growth rate produced by the gravitational instability of a layer of buoyant fluid each depend on the cube root of its viscosity. In the light of these differing results, we conclude that the geometry of the buoyant region of partial melt must be considered when developing gravitational instability models of island arc and mid-ocean ridge volcanism. We also conclude that if the expressions for a cylindrical region are appropriate then the observed spacing of volcanoes cannot be used to infer the viscosity and hence the melt fraction of the source region. A fixed-flux release of fluid was achieved by injecting buoyant fluid through a narrow slot in the base of a large tank of denser fluid. The interface between the two fluids was found to be unstable and diapirs were formed; however, the rising diapirs were followed by a continuous, uniform sheet of buoyant fluid. We suggest, therefore, that thermal or rheological effects need to be included in fluid-mechanical models if volcanism which is periodic in position and time is to be explained.