Depending on the orientation of thermal and compositional gradients at the margins of a cooling intrusion, double-diffusive convection may take one of two principal forms. If the body is cooling at a steep wall, so that both gradients are horizontal, the result can be boundary layer flow up or down the wall and segregation of the fractionated liquid under the roof or on the floor. If cooling is through the roof or floor of a broad sheet-like body, the gradients will be vertical and if the physical properties of the liquid are appropriate, they could lead to broad horizontally stratified convection. The first case has been analysed by Nilson et al. (1985) and shown to be a physically plausible explanation for certain types of zoned magmas, but the second case, in which both gradients are vertical, appears to be less likely, especially in slowly cooled intrusions in which temperatures are controlled by crystal-liquid equilibrium. Depending on whether the magma is losing heat through its roof or floor and on whether its density increases or decreases with cooling and differentiation, four possible conditions are possible. A linear stability analysis based on the experimental and theoretical criteria of Baines & Gill (1969) and Shirtcliffe (1969) indicates that neither calc-alkaline liquids that accumulate under their roof nor tholeiitic liquids ponded on their floor are likely to convect in stratified layers. Even though the gradients and relative diffusivities of heat and composition may be in a direction that tends to destabilize the magma, the magnitude of the compositional effect on density is so great that the temperature gradient would have to diverge widely from the liquidus, and convection would be possible only with large amounts of under-cooling or superheat. A similar mechanism, based on opposed gradients of two chemical components of differing diffusivities, may provide an alternative explanation for certain types of stratification, but field relations in the Skaergaard Intrusion do not support any origin that is primarily dependent on gravity.
No takes yet. Share an insight, caveat, or question.
Alexander R. McBirney (1985) studied this question.