As basaltic rocks formed at the base of the oceanic floor are transported back to the Earth's mantle along subduction zones, they undergo transitions and introduce compositional and thermal heterogeneities in the deeper parts of the mantle. Studying the melting phase relations of basaltic lithologies at elevated pressures and temperatures provides insights into what potentially happens at different depths in the lower mantle now and throughout the past billion years of active plate tectonics. Using laser heateddiamond anvil cell experiments combined with in situ X-Ray Diffraction measurements at synchrotron sources, we revisit the crystallization and melting properties of natural basaltic samples at 60-100 GPa and up to 4000 K. Diffraction patterns highlight the major phases: bridgmanite and Ca-perovskite, followed by crystallization of Si-rich phases (mainly stishovite) and Calcium Ferrite (CF-type) Na and Al-rich phase. Recovered samples were prepared using focused ion beam techniques for detailed chemical analyses of the extracted thin sections by electron microscopies in order to resolve sub-micron features and understand the chemical partitioning of elements induced by melting at high pressure and temperature conditions. We confirm that the liquidus phase is Ca-perovskite, which segregates during melting and is recovered as rings that encapsulate a melt pool throughout the studied pressure range. The melt pocket shows a concentric structure consisting of an alumino-silicate envelope surrounding an Fe-rich silicate part. At the center of samples, an Fe-O-S metal pond is often observed. We associate the observation of segregation of liquid phases to capillary forces. The differentiation of melt pockets into three melts is tentatively attributed to Marangoni effects, i.e. temperature-induced surface tension gradients in the samples. Central metal ponds are indirectly best interpreted as related to the disproportionation reaction of Fe 2+ into Fe 3+ and Fe (0) in bridgmanite whereas the two silicate-melt pools could be associated to the formation of two immiscible liquids upon melting of basalts. On the basis of these observations, we propose that melting of basaltic lithologies at lower mantle pressures could lead to important chemical differentiation mostly characterized by Fe enrichment at increasing depth.