Testing for compositional convection in silicate melts : crystal growth experiments and a petrographic study of a differentiated ring dyke

Convection in magma chambers, driven by compositional density differences, is thought to be a major process in fractionation of liquid from crystals in the differentiation of magmas. Compositional convection has been tested for by undertaking two sets of experiments in silicate melts, and by re-examining the vertically differentiated gabbro-granophyre Glen More ring dyke on the Isle of Mull. Crystal growth quenching experiments have been carried out in a synthetic basalt in which iron is replaced by cobalt. Co-Mg olivine was overgrown on olivine seeds cemented in alumina crucibles and the glasses produced on quenching were analysed by EPMA for compositional variation at and above the overgrowth-glass interface. Boundary layers up to 50?m wide, and depleted in Co and Mg by upto 25 %, have been found at crystal-glass interfaces, and plumes of the same melt have been detected above the apexes of growing olivine crystals. The computed density difference which causes the convection of this boundary layer, melt is in the region of 1 %. This phenomenon, known as compositional convection, has been seen associated with dissolving crystals in silicate melts, and around growing crystals in aqueous salt solutions, but this is the first time it has been reported accompanying crystal growth in silicate melts. Other experiments in which hematite underwent dissolution in a natural basalt melt show that dense, Fe-enriched melt can be lifted above its origin by the buoyant rise of bubbles. In both sets of experiments a zone of buoyant boundary layer melt has been produced by side-wall crystallization of hercynite. This zone of melt has convected up and has ponded beneath the meniscus by a process that can be likened to side-wall crystallization in magma chambers. Petrological, mineralogical, textural, and new geochemical evidence from the re-examination of the Glen More ring dyke strongly suggests that the petrological variation is the result of a magma mixing mechanism. It is proposed that this was produced by the injection of a basic magma which underwent partial consolidation. This was then followed by a second injection of a silicic magma which underwent partial mixing with the dioritic residual magma remaining after crystallization of gabbros from the initial magma