A potential driving force for cumulate compaction and melt segregation in magmas arises from the change in crystal solubility with temperature. Spatial thermal variation in a cumulus crystal pile sets up a gradient in interstitial melt composition, which in turn provides a diffusion potential for silicate liquid species. Mass transport in response to this potential, referred to as thermal migration, drives adcumulus mineral growth in the cooler regions of a cumulus crystal pile and segregation of interstitial melt toward warmer regions. Heat loss through the cumulate mush zone to the country rock promotes expulsion of intercumulus liquid back toward the magma reservoir. Heat loss from the mush to cooler convective draughts within the chamber promotes trapping of intercumulus liquid. Liquid state Soret diffusion alters this mechanism only in minor chemical detail except when Soret fractionation produces changes in liquid chemistry that exceed those produced by crystal fractionation. This can cause solubility curves to be intersected at high temperature rather than low T, burying the intercumulus liquid in boundary layer and preventing its escape back into the magma.

Quartz saturation in silicic systems may be affected in this way. We demonstrate this process with laboratory experiments in the systems Mg2SiO4-SIO2 and mid-ocean ridge basalt. Cumulate compaction rates by thermal migration are limited by diffusion rate, but length scales for compaction increase linearly in time as long as the saturation conditions and temperature are unchanged. Adcumulus crystal growth remote from the accumulation front by thermal migration, although locally at equilibrium with the intercumulus melt, is not in chemical equilibrium with the main magma reservoir. The operation of thermal migration can reconcile two commonly observed, but seemingly contradictory, features of adcumulate rocks: unzoned cumulus crystals (reflecting isothermal crystallization at one place) within cryptically varying mesoscopic layers (reflecting spatial variations of temperature). The chemical details of this process are complicated by selective interdiffusion of components in the intercumulus liquid along compositional/thermal gradients. Compatible and incompatible elements show anomalous enrichments, thus complicating estimates of residual intercumulus liquid porosity and magma composition from cumulus rock chemistry. Thermal migration does not operate independently of magma body cooling or to the exclusion of the other processes of cumulate crystal compaction. Physically, the most promising opportunities for thermal migration to produce observable results occur within individual layers of large layered sequences. Large mafic intrusions cool slowly enough that mass diffusion in the intercumulus melt can observably rearrange constitutents of individual layers of cumulus crystal pile in chemistry, modal mineralogy, and texture. Some of the expected consequences of the process are found in layered sequences of mafic intrusions.