Crystallization in a double‐diffusive system

Abstract

The experiments reported here are a first attempt to model in a simple laboratory system some of the physical processes occurring during crystallization in magma chambers and to examine possible effects of different molecular diffusion rates on the formation of layered igneous rocks. Our laboratory results suggest that stratification in the melt and ‘double‐diffusive’effects during solidification of the magma should be taken into account when developing theories of the layering. Explicitly, we have extended earlier experiments on double‐diffusive convection, which show how layers may readily form from smooth gradients when two components with different molecular diffusivities have opposing effects on the density gradient. The new experiments included the influence of crystallization and have used an aqueous solution Of Na2CO3 cooled in various configurations. When cooling was from the top, the crystals formed in horizontal layers, whether the initial fluid was homogeneous, was stratified with a constant gradient, or was stratified in several distinct layers. The dominant effect was the formation of a cold but light fluid layer against the cooled upper boundary as the denser crystals grew. The crystals were small and closely packed in the upper layer (which was vigorously converting) and larger and more loosely packed in the lower, more quiescent fluid. In the case where there were preexisting layers there was an abrupt increase in the growth rate of crystals as they crossed an interface. With sidewall cooling and a constant gradient the lighter solution left behind after crystallization streamed upward in a thin boundary layer right to the top of the tank. The outer flow consisted of circulation in nearly horizontal converting layers, each of which was depressed slowly as the light fluid collected at the top. When a constant gradient of Na2CO3 was cooled from below, crystallization produced a ‘finger’ instability, leading to the growth of a mixed layer in the fluid above. An experiment using opposing gradients of Na2CO3 and K2CO3 (with Na2CO3 rich on top) showed that crystals settling to their own density level and then redissolving can reverse the original sense of the gradient and produce small‐scale double‐diffusive layering in the liquid.