The processes that control leucosome compositions in metasedimentary granulites: perspectives from the Southern Marginal Zone migmatites,Limpopo Belt,South Africa

Anatexis of metapelitic rocks at the Bandelierkop Quarry (BQ) locality in the Southern Marginal Zone of the Limpopo Belt occurred via muscovite and biotite breakdown reactions which, in order of increasing temperature, can be modelled as: (1) Muscovite + quartz + plagioclase = sillimanite + melt; (2) Biotite + sillimanite + quartz + plagioclase = garnet + melt; (3) Biotite + quartz + plagioclase = orthopyroxene ± cordierite ± garnet + melt. Reactions 1 and 2 produced stromatic leucosomes, which underwent solid‐state deformation before the formation of undeformed nebulitic leucosomes by reaction 3. The zircon U–Pb ages for both leucosomes are within error identical. Thus, the melt or magma formed by the first two reactions segregated and formed mechanically solid stromatic veins whilst temperature was increasing. As might be predicted from the deformational history and sequence of melting reactions, the compositions of the stromatic leucosomes depart markedly from those of melts from metapelitic sources. Despite having similar Si contents to melts, the leucosomes are strongly K‐depleted, have Ca:Na ratios similar to the residua from which their magmas segregated and are characterized by a strong positive Eu anomaly, whilst the associated residua has no pronounced Eu anomaly. In addition, within the leucosomes and their wall rocks, peritectic garnet and orthopyroxene are very well preserved. This collective evidence suggests that melt loss from the stromatic leucosome structures whilst the rocks were still undergoing heating is the dominant process that shaped the chemistry of these leucosomes and produced solid leucosomes. Two alternative scenarios are evaluated as generalized petrogenetic models for producing Si‐rich, yet markedly K‐depleted and Ca‐enriched leucosomes from metapelitic sources. The first process involves the mechanical concentration of entrained peritectic plagioclase and garnet in the leucosomes. In this scenario, the volume of quartz in the leucosome must reflect the remaining melt fraction with resultant positive correlation between Si and K in the leucosomes. No such correlation exists in the BQ leucosomes and in similar leucosomes from elsewhere. Consequently, we suggest disequilibrium congruent melting of plagioclase in the source and consequential crystallization of peritectic plagioclase in the melt transfer and accumulation structures rather than at the sites of biotite melting. This induces co‐precipitation of quartz in the structures by increasing SiO₂ content of the melt. This process is characterized by an absence of plagioclase‐induced fractionation of Eu on melting, and the formation of Eu‐enriched, quartz + plagioclase + garnet leucosomes. From these findings, we argue that melt leaves the source rapidly and that the leucosomes form incrementally as melt or magma leaving the source dumps its disequilibrium Ca load, as well as quartz and entrained ferromagnesian peritectic minerals, in sites of magma accumulation and escape. This is consistent with evidence from S‐type granites suggesting rapid magma transfer from source to high level plutons. These findings also suggest that leucosomes of this type should be regarded as constituting part of the residuum from partial melting.