Compositional and thermal convection in magma chambers

Abstract

Magma chambers cool and crystallize at a rate determined by the heat flux from the chamber. The heat is lost predominantly through the roof, whereas crystallization takes place mainly at the floor. Both processes provide destabilizing buoyancy fluxes which drive highly unsteady, chaotic convection in the magma. Even at the lowest cooling rates the thermal Rayleigh number Ra is found to be extremely large for both mafic and granitic magmas. Since the compositional and thermal buoyancy fluxes are directly related it can be shown that the compositional Rayleigh number Rs (and therefore a total Rayleigh number) is very much greater than Ra. In the case of basaltic melt crystallizing olivine Rs is up to 10⁶ times greater than Ra. However compositional and thermal buoyancy fluxes are roughly equal. Therefore thermal and compositional density gradients contribute equally to convection velocities in the interior of the magma. Effects of thermal buoyancy generated by latent heat release at the floor are included.

The latent heat boundary layer at the floor of a basaltic chamber is shown to be of the order of 1 m thick with very low thermal gradients whereas the compositional boundary layer is about 1 cm thick with large compositional gradients. As a consequence, the variation in the degree of supercooling in front of the crystal-liquid interface is dominated by compositional effects. The habit and composition of the growing crystals is also controlled by the nature of the compositional boundary layer. Elongate crystals are predicted to form when the thickness of the compositional boundary layer is small compared with the crystal size (as in laboratory experiments with aqueous solutions). In contrast, equant crystals form when the boundary layer is thicker than the crystals (as in most magma chambers). Instability of the boundary layer in the latter case gives rise to zoning within crystals. Diffusion of compatible trace elements through the boundary layer can also explain an inverse correlation, observed in layered intrusions, between Ni concentration in olivine and the proportion of Ni-bearing phases in the crystallizing assemblage.