CONVECTION IN AID OF ADCUMULUS GROWTH

Abstract

Sheet cooling occurs for most large magma bodies emplaced into the earths crust The convection that ensues is driven by the stove effect of the feeder, by cooling principally at the roof, and to an important degree by two-phase flow of crystal-liquid suspensions accelerated by rapid crystal growth upon compression. Significant floor cooling through the cumulate substrate is limited to the early history of an intrusion. Thereafter a thermal maximum in the T-Z profile (a latent-heat hump) always occurs at the cumulus interface. If crystals grow, the latent heat is carried away chiefly by the magma. The solute rejected by the growing interface is buoyant for mafic and ultramafic cumulates but dense for all felsic and feldspar-cotectic cumulates not containing cumulus iron oxide minerals. The rejected solute (RS) is light for calc-alkaline cumulates in general, a fact probably germane to the origin of rhyolite by sidewall crystallization. Near the cumulate interface, the ratio Rp of compositional to thermal effects on liquid density lies in the range 10–106, indicating a strong compositional control on density that cannot be overcome by available thermal contrasts. The solidification of the cumulate is enhanced by upward removal of light RS but impeded by stagnation of dense RS. The existence of felsic flat floor adcumulates proves that adcumulus growth can occur largely by diffusion, as all other options are precluded. This result is entirely consistent with known diffusivities of heat and mass, for accumulation rates near ? cm yr?1. When higher accumulation rates occur, some residual liquid is trapped. The flux ratio of heat to matter during adcumulus growth is about 300 cal g?1. From the flux ratios and the one-dimensional estimates of cooling rates, typical values of the thermal and compositional gradients during adcumulus growth can be obtained. After a new batch of supercooled magma arrives the thermal gradient drops from near infinity to > 5000? km?1 during nucleation and then to a steady state near 40°C km?1 at the interface during adcumulus growth. The latent heat hump is paired with another thermal maximum above the boundary layer, which is therefore a heat sink. By consideration of the thermal structure and the buoyancy of rejected solute it is determined that double diffusive convection is inconsistent with the adcumulus growth of any type of floor cumulate, and with any growth of an ultramafic cumulate. Infiltration metasomatism is an orthocumulus process rather than an adcumulus one, and it produces mesocumulates. Multiple stratification of magma will not arise from growth at the floor, but storage of buoyant RS in a polymerized layer rich in plagioclase component appears to occur on a rapid timescale in snowflake troctolite and may have led to anorthosite formation on a slow timescale in the Stillwater Complex. Magma stratification by multiple injection is likely to be unstable to two-phase convection. Compaction of cumulates is limited to cases where sufficiently thick crystal mush can be shown to exist, and such thicknesses are rare in large intrusions. Compaction is therefore not a general alternative to adcumulus growth. Cumulate theory is very much alive and able to predict testable ideas about silicate diffusivities and convection.