When the subcontinental lithospheric mantle undergoes heating and/or extension, some of the earliest mafic melts to be generated are those rich in volatiles and potassium. In some cases, e.g. when a plume impinges on thick cratonic lithosphere or when the amount of extension is very small, K-rich mafic igneous racks may be the only surface expression of mantle melting. The Alto Paranaiba Igneous Province, in SE Brazil, is one of the world's most voluminous mafic potassic provinces (> 15000 km(3)), which until recently was relatively unknown. The magmas were emplaced into a narrow Proterozoic mobile belt close to the surface margin of the Sao Francisco craton, and it is one of several Cretaceous alkaline igneous provinces that are located around the margin of the Parana' sedimentary basin in Brazil and Paraguay. Detailed geochemical analyses of samples from throughout the Alto Paranaiba Igneous Province show that it is composed of a relatively diverse suite of ultrapotassic-potassic, ultramafic-mafic, silica-undersaturated lavas and hypabyssal intrusions, i.e. kimberlites, madupitic olivine lamproites and kamafugitic rocks. These all have very high concentrations of incompatible trace elements and are all strongly enriched in light rare earth relative to heavy rare earth elements (e.g. La/Yb = 50-230). Wide variations in major element ratios, which are unrelated to the effects of crystal fractionation in these magmas (e.g. CaO/ Al2O3), suggest that the mafic potassic rocks were derived from a heterogeneous mantle source. They show relatively restricted ranges of initial Sr-87/Sr-86 (0.70436-0.70588) and epsilon(Nd85), values of -4 to -8, intermediate between Group I and II South African kimberlites. T-DM Nd isotope model ages of similar to 900 Ma suggest that the magmas were derived by the remobilization of subcontinental lithospheric mantle that had been enriched by small-volume K-rich melt fractions since the Late Proterozoic. New K/Ar ages for mica separates show that the kimberlites, madupitic olivine lamproites and kamafugitic rocks were emplaced together with large carbonatite-bearing plutonic complexes at similar to 85 Ma. Reconstructions of plate motions show that, at this time, the location of the Alto Paranaiba Igneous Province coincided with the postulated position of the present-day Trindade (or Martin Vaz) plume. We propose that the widespread Late Cretaceous alkaline magmatism in SE Brazil may have been caused by impingement of this plume on the base of the subcontinental lithosphere. Heat penetrating the lithosphere, both by conduction and advection by asthenospheric-source decompression melts, may have caused melting of the readily fusible parts of the lithospheric mantle and the genesis of mafic potassic and (after fractionation) carbonatite magmas. The Proterozoic mobile belt (the Brasilia Belt) appears to have acted as a 'thinspot' relative to the adjacent Sao Francisco craton, allowing greater upwelling and melting of the asthenosphere. Subsequently, as the craton passed over the plume, volcanism was 'switched off' until the Early Tertiary when the plume re-emerged from beneath the westward drifting South America continent and was the magma source for oceanic-islands and seamounts of the Trindade Vitoria chain.
