Kilauea Iki lava lake formed in 1959 as a closed chamber of 40 million m³ of picritic magma. Repeated drilling and sampling of the lake allows recognition of processes of magmatic differentiation, and places time restrictions on the periods when they operated. This paper focuses on evidence for the occurrence of lateral convection in the olivine-depleted layer, and constraints on the timing of this process, as documented by chemical, petrographic and thermal data on drill core from the lake.

Lateral convection appears to have occurred in two distinct layers within the most olivine-poor part of the lake, created a slightly olivine-enriched septum in the center of the olivine-depleted section. A critical marker for this process is the occurrence of loose clusters of augite microphenocrysts, which are confined to the upper half of the olivine-poor zone. This process, which took place between late 1962 and mid-1964, is inferred to be double-diffusive convection.

Both this convection and a process of buoyant upwelling of minimum-density liquid from deep within the lake (Helz, R.T., Kirschenbaum H. and Marinenko, J.W., 1989. Diapiric melt transfer: a quick, efficient process of igneous differentiation: Geological Society of America Bulletin, v. 101, 578–594) result from the fact that melt density in Kilauea Iki compositions decreases as olivine and augite crystallize, above the incoming of plagioclase. The resulting density vs. depth profile creates (1) a region of gravitationally stable melt at the top of the chamber (the locus of double-diffusive convection) and (2) a region of gravitationally unstable melt at the base of the melt column (the source of upwelling minimum-density melt, Helz, R.T., Kirschenbaum H. and Marinenko, J.W., 1989. Diapiric melt transfer: a quick, efficient process of igneous differentiation: Geological Society of America Bulletin, v. 101, 578–594). By contrast the variation of melt density with temperature for the 1965 Makaopuhi lava lake does not show a decrease in density as temperature decreases, so neither process should have occurred in that lava lake. Because many mafic magmas crystallize significant olivine and/or pyroxene before they begin to crystallize plagioclase, the density relations observed for Kilauea Iki, and the processes that result from them, may be relevant to crystallization in other mafic magma chambers. The results for the 1965 Makaopuhi lava lake emphasize the role of bulk composition as a critical control on magmatic processes.