Topology of syntectonic melt-flow networks in the deep crust: Inferences from three-dimensional images of leucosome geometry in migmatites

We have determined the three-dimensional form of leucosome in two migmatite s produced by syntectonic anatexis of different protoliths: (1) stromatic m igmatite derived from pelite, which comprises sheets of leucosome (quartzof eldspathic layers with Grt) with walls of melanosome (Bt-rich selvages) in mesosome (schistose layers of Bt + Pl + Qtz +/- Grt +/- Crd) and (2) a migm atitic garnet amphibolite derived from basalt, which is composed of spindle -shaped leucosomes (Qtz + Pi), spatially associated with peritectic Grt, in melanosome (Hbl + Qtz +/- Cpx). Three-dimensional images were generated fr om two-dimensional representations of spatial data obtained by two methods: (1) serial grinding and (2) high-resolution X-ray computed tomography (HR X-ray CT). Projections of three-dimensional images of stromatic migmatite d erived using data from either method show the planar nature of leucosome th roughout the sample; melt transport through this rock when it was partially molten could be modeled as flow in parallel conduits. In the image derived from HR X-ray CT data, garnet in leucosome is only rarely in contact with melanosome, which suggests these garnet grains were suspended in melt durin g flow. Projections of three-dimensional images of the migmatitic garnet-am phibolite do not reveal the full extent of leucosome connectivity, due to t he irregular geometry of leucosome. Connectivity in this sample can be show n, however, by virtual slicing of the three-dimensional images perpendicula r to the plane of the two-dimensional representations (approximately parall el to the lineation defined by the leucosome), and by using three-dimension al projections of a single leucosome connectivity "tree" constructed by pro jecting leucosome patches from slice to slice and noting the overlap. Based on leucosome geometry and volume, we estimate effective porosity for flow in this rock to have been 20 vol% at stagnation. Leucosome in the migmatiti c garnet-amphibolite occurs in strain shadows around garnet, which are infe rred to have been obstacles to flow along linear paths. Blocking of inferre d flow channels by garnet contributes to the high degree of pathlength tort uosity in this sample (tau = 2-6), which is expressed visually by the compl ex form of the leucosome in three dimensions. Cross-sectional areas for ind ividual inferred melt flow paths are highly variable lover 2-3 orders of ma gnitude) and minimum channel radius is changeable (by similar to 1 order of magnitude), meaning there was large variability along the channels and imp lying strong local flow divergences. Based on these data, unusually straigh t and uniform channels would have dominated the mesoscopic melt flux throug h this rock when it was partially molten.