The Corsican batholith, formed during the Variscan orogeny, was studied with particular.emphasis on the respective roles played by crustal- and mantle-derived material in granite genesis. Major- and trace-element data identify: the various types of magma and their genetic relationships; the magmatic processes that gave rise to the observed rocks; and the primitive liquids. O-Sr-Nd-isotope systematics constrain the origin of primitive liquids. Trace-element models provide data on the various source materials and their possible geodynamic setting. The evolution of two main calc-alkaline plutonic associations is compared: an older high-Mg-K association (U1) with both ultrapotassic mafic and silicic rocks, the latter ranging from monzonite to leucosyenogranite, and a younger calc-alkaline composite association (U2) involving a mafic cumulate sequence and granodiorite to leucomonzogranite. Both types of granite are accompanied by mafic units that were non-cogenetic with surrounding granitic melts. Trace-element calculations indicate that a primitive liquid of monzodioritic composition gave rise to the U1 Mg-K granite by fractional crystallization, amphibole and titanite playing a major role, which, with increasing crystallization, gave decreasing REE contents without a strong negative Eu anomaly. The U2 calc-alkaline suite was formed by fractional crystallization involving feldspar and LREE-rich minerals, i.e. monazite, from a liquid of monzogranitic composition. Field and petrographical data identify magma mingling in both U1 and U2, between mafic and silicic rocks, but geochemical data only indicate how magma-mixing processes led to U2 granodiorite. Geochemical modelling shows that a single protolith of calculated graywacke composition yielded the two associations under different melting conditions: the high-Mg-K monzodiorite melt was formed after partial melting of a 30% non-modal batch of granulite-facies metamorphic protolith (low p(H2O)), but the calc-alkaline monzogranite was formed by the same process in an amphibolite-facies source of similar composition (higher p(H2O)). The primitive granite magmas of both associations show the same crustal characteristics, i.e.: Sr(i) = 0. 706-0.707; epsilon(Nd) (t) = -4.3 to -2.2 and deltaO-18 = +7 to +8 parts per thousand. This provides evidence that the composition of both U1 and U2 granite melts probably was controlled by physico-chemical fusion rather than protolith-composition parameters. The U1 ultra-potassic mafic rock is thought to be of mantle origin, i.e. a deep source containing phlogopite (+/- garnet). Zone refining led to a significant increase of incompatible trace-element contents during ascent of the magma. Mantle-crust interaction lowered the La/Yb and Nd-143/Nd-144 ratios, but the Sr-87/Sr-86 ratio increased. Nevertheless, interaction with associated U1 Mg-K granite was not important and the crustal component is thought to be different from the associated granitic magma. In U2, the mafic cumulate sequence, clearly of mantle origin, has E-MORB characteristics and seems the result of 10% non-modal partial melting of a heterogeneous mantle of spinel or amphibole peridotite without garnet. Some interaction occurred with the associated granitic melt. The E-MORB character of the mafic rock indicates that a large part of the batholith. and at least the U2 magmatic association, was generated under post-collisional extensional conditions.
