We report on the occurrence, petrology and geochemistry of recently recognized leucocratic plagioclase-rich microgranular inclusions hosted by two granite facies in the late-Caledonian Galway Granite, Connemara, Ireland. They have been recorded at 66 localities along an ESE trending, 4 km wide corridor which incorporates the contact zone between their host granites (i.e. The Megacrystic Granite and the Mingling and Mixing Zone Granodiorite). The inclusions are discoidal in shape and oriented parallel to the general ESE trending foliation in the granites with the most elongate (6.0 × 0.6 cm) occurring in zones of strongest fabric intensity. Contacts between the inclusions and the host granite are sharp with no chilled margin visible. They display a fine-grained (<1 mm) interlocking texture with occasional crystals of plagioclase ranging up to 2 mm in length. Microprobe analysis shows that the plagioclase is essentially oligoclase (An22–32) in composition and is similar to that (i.e. An21–30) occurring in the host granites. Furthermore, the oligoclase accounts for between 61 and 78% of the mode which is reflected in the major element chemistry of the inclusions. Other minerals (in decending order of abundance) include K-feldspar, quartz, biotite and magnetite. The origin of the inclusions is unclear. However, the results of the microprobe analysis provide evidence of a link between them and their host granites.

The distribution of melt has been mapped in a granite pluton deformed whilst it contained melt. At the outcrop scale, leucomonzogranite melt was segregated from hornblende monzogranite when the rigid crystal framework was tectonically compacted. The melt collected in well-defined, structurally controlled sites that formed during dextral, non-coaxial, strike-slip shearing. The segregations are generally isolated, but locally they link to form extensive branched arrays which drained larger volumes of granite in a two-step process. First, melt drained from the compacting matrix through the array and pooled along dilatant foliation planes; later, the melt moved farther away when a single planar melt-transfer channel formed.

Thin section maps show that most melt was distributed in the foliation plane and along the lineation in the crystallising matrix. The location of melt at the grain scale is primarily controlled by the feldspar-dominated shape fabric of the crystal framework, and not by tectonic stresses as at the outcrop scale. Tectonic stresses account for the relatively small proportion of melt films located in grain boundaries normal to the lineation. The distribution of melt-bearing grain boundaries outlines larger domains in the thin sections that form a linked three-dimensional network through which melt moved within the crystallising framework.

Dehydration melting of crustal rocks may commonly occur in response to the intrusion of mafic magma in the mid- or lower crust. However, the relative importance of melt buoyancy, shear or dyking in melt generation and extraction under geologically relevant conditions is not well understood. A numerical model of the partial melting of a metapelite is presented and the model results are compared with the Ivrea-Verbano Zone in northern Italy. The numerical model uses the mixture theory approach to modelling simultaneous convection and phase change and includes special ramping and switching functions to accommodate the rheology of crystal-melt mixtures in accordance with the results of deformation experiments. The model explicitly includes both porous media flow and thermally and compositionally driven bulk convection of a restitecharged melt mass. A range of melt viscosity and critical melt fraction models is considered. General agreement was found between predicted positions of isopleths and those from the Ivrea-Verbano Zone. Maximum melt velocities in the region of porous flow are found to be 1 × 10−7 and 1 × 10−1m per year in the region of viscous flow. The results indicate that melt buoyancy alone may not be a sufficient agent for melt extraction and that extensive, vigorous convection of partially molten rocks above mafic bodies is unlikely, in accord with direct geological examples.

Fluid dynamic modelling of crystallising calc-alkalic magma bodies has predicted that differentiated liquids will ascend as boundary layers and that accumulation of these buoyant liquids near chamber roofs will result in compositionally stratified magma chambers. This paper reports physical features in La Gloria Pluton that can be interpreted as trapped ascending differentiated liquids. Leucogranitic layers decimetres thick, which are locally stratified, are trapped beneath overhanging wall contacts. The same felsic magmas were also preserved where they were injected into the wall rocks as dykes and as large sill complexes. These rocks do not represent differentiated magmas produced by crystallisation along the exposed walls because the felsic layers occur at the wall rock contact, not inboard of it. Rather, we speculate that evolved felsic liquids are generated by crystallisation all across the deep levels of chambers and that initial melt segregation occurs by flowage of melt into tension fractures. Melt bodies so formed may be large enough to have significant ascent velocities as diapirs and/or dykes. The other way in which the leucogranite occurrence is at variance with the convective fractionation model is that the ascending liquids did not feed a highly differentiated cap to the chamber, as the composition at the roof, although the most felsic in this vertically and concentrically zoned pluton, is considerably more mafic than the trapped leucogranitic liquids. This suggests that these evolved liquids were usually mixed back into the main body of the chamber. Backmixing may be general in continental-margin calc-alkalic magmatic systems, which, in contrast to those in intracontinental settings, rarely produce volcanic rocks more silicic than rhyodacite. That the highly differentiated liquids are preserved at all at La Gloria is a result of the unusual stepped nature of the contact and the entirely passive mode of emplacement of the pluton, which, in contrast to ballooning in place, does not result in wall zones being “scoured”.