Compositional convection in a reactive crystalline mush and melt differentiation

Fractional crystallization and partial melting involve relative motion of liquid and solid phases and chemical and thermal interactions between them. To elucidate some physical principles of thermosolutal convection in a reactive porous medium, we describe experiments on the directional solidification of aqueous ammonium chloride solutions. The addition of small amounts of a polymerizing agent permits variation of the solution viscosity independently of thermal conditions, the phase diagram, and permeability. Solutions were cooled from below, and crystallization developed at the base of the tank generating at first a field of thin plumes of light residual fluid, released from the boundary layer at the top of the “mush.” The interstitial fluid within the mush also became unstable eventually, the onset of convection occurring when the porous medium Rayleigh number of the mush reached a critical value. This threshold value was found to be 25 at low initial superheat and to decrease with increasing superheat. Local thermodynamic equilibrium between crystals and liquid within the mush coupled the evolution of temperature, composition, and porosity. Convective motions locally caused dissolution and precipitation, and hence fluctuations of porosity developed. Dissolution occurred preferentially in the central parts of upwellings, and the upflow gradually focused, being ultimately channelized into narrow “chimneys” devoid of crystals. All else being equal, the areal density of chimneys was greater the lower the viscosity. Chimney diameter increased with increasing solution viscosity. In the liquid above the growing mush, the convective plumes were very similar to salt fingers. Depending on solution viscosity and temperature gradient, they exhibited a phenomenon of collective instability such that vertical motion was disrupted by wave instabilities. The base plate temperature chosen was above the eutectic, and hence the total amount of crystals at the end of experiments with the same initial composition and the same final temperature was a constant fixed by the phase diagram. The spatial distribution of crystals and the final porosity of the mush were, however, determined by the strength of compositional convection as measured by the porous medium Rayleigh number. When no convection occurred, mush thickness finally became equal to the initial layer thickness, and the system was homogeneous. When compositional convection occurred, the overlying reservoir underwent chemical evolution, and mush growth slowed dramatically. In experiments with progressively lower solution viscosity (and hence more vigorous convection), final mush thickness was progressively less and final porosity lower. Final mush thickness was found to scale with solution viscosity to the power +0.33. During fractional crystallization of magma the effects of compositional convection can be recorded in the chemical and mineralogical features of cumulate rocks. We speculate that fossil chimney structures can be found in the Lower Zone of the Bushveld ultramafic complex in the form of iron‐rich, platinum‐bearing dunite pipes. This could explain the adcumulate nature of the Lower Zone rocks and the chemical and mineralogical similarities between the pipes and the overlying Merensky Reef; the Merensky Reef would be interpreted as a hiatus in mush growth. The results may also have applications to the flow structure in regions of partial melting and in the Earth's core.