Melt percolation, concentration and dyking in the Hawaiian mantle plume and overriding lithosphere: links to the evolution of lava composition along the volcanic chain

International audienceOceanic island basalts and related magmatic rocks from Hawaii are derived from a compositionally heterogeneous mantle plume. Here we describe how this heterogeneity results from the transport of filaments of a specific composition in the plume, representing a relatively small volume of rocks (~15%) interbedded inside a dry peridotite mantle. Four types of filaments are considered: sub-primitive mantle, ultra-low-velocity zone, fertilized-harzburgite and eclogite type filaments. We present a model that describes the flow within a plume and the stress field in the overriding viscoelastic lithosphere and that can determine, from depth to the surface, the melting rate, composition and trajectory of melts produced within each type of filament. Our model shows that (i) the filaments melt at a depth greater than 5 GPa, where the temperature gap between the solidus and liquidus is narrow (~ 40°C to 80°C), and (ii) the volume of filaments is small relative to the total volume of mantle, which therefore allows the latent heat required for the partial melting to be provided via conduction inside the hot plume. The primitive melts produced inside the filaments, occasionally mixed with the melt derived from an eclogite filament, represent a volume comparable with that expected in a plume composed only of dry peridotite that partially melts to a degree of ~10% at the interface between the spinel and garnet fields (60–70 km depth). In particular, in the centre of the plume, sub-primitive mantle filaments produce up to 30% tholeiite-picrite melts, whereas in fertilized-harzburgite filaments, the mantle melts completely to produce a melt having a meimechite-like composition. A key finding is that the fractional crystallization of these melts likely forms the so-called “primary mantle–derived alkaline magmas” along with dunites and olivine-rich cumulates. Our plume model shows that the mantle flow divides into two parts. The first corresponds to hot flowlines that originate at a depth of ~200 km and at a distance of less than 25 km from the plume axis. Along these flowlines, when the mantle reaches a pressure of 5 GPa, the partially molten horizon in filaments is sufficiently thick for the interstitial melt to be squeezed out via dykes. This melt eventually ponds as sills in a subrectangular zone that is located inside the overlying lithosphere, between 70 and 50 km depth and centred over a distance of less than 40 km on either side of the axis. This zone is designated as the shield magmatic reservoir. The volatile-rich melt inside the sills infiltrates the surrounding mantle lithosphere and partially melts it. After ~0.1 Ma, the melt resumes its vertical ascent via dykes and eventually ponds and differentiates within subcrustal magma chambers located below active shield volcanoes. This sequence of processes matches the expected volume, petrology and geochemistry recorded for shield volcanoes. The second part of the melt flow does not pond within the shield magmatic reservoir. Rather, the mantle cold flowlines, originating at ~200 km depth and at 25 to 35 km from the plume axis, discharge their interstitial melt through dykes that were initially generated deeper, at ~5 GPa. The melt reaches the Moho at 100 to 150 km from the plume axis, where it forms magmatic bodies within which the melt differentiates. This melt likely represents that observed in pre- and postshield volcanoes. Finally, at ~70 km from the plume axis and at a depth greater than 200 km, the flowlines are subvertical. They then deflect at ~180 km depth and rotate toward the horizontal and eventually transit at 10° to 20° to the horizontal across a ~200 km distance from the axis and reach ~140 km depth. The fertilized-harzburgite and sub-primitive mantle/ultra-low-velocity zone filaments that flow along these elbows partially melt by a few to several percent. The resulting interstitial melt has a kimberlite-like composition. Thereafter, the excess pressure at the top of the filament at ~200 km from the axis overcomes the threshold for dyking and thus allows the escape of the interstitial melt via dykes ponding in subcrustal magma chambers or emerging directly at the surface. These melts have a composition similar to that associated with rejuvenated volcanism. We use the nature and the composition of whole erupted magmas and the seismic structure along the Hawaiian chain to validate this model