In subduction zones, melting and dehydration of the subducted slab introduce material into the mantle wedge and modify its chemical and isotopic composition. As a consequence, island arc lavas differ significantly from mid-ocean ridge basalts and ocean island basalts. In some arcs, the composition of lavas is strongly influenced by the sedimentary material introduced with the slab; in others, magma composition is mainly affected by aqueous fluids released by the slab. The Lesser Antilles arc is known for its extreme continental-crust-like signature but for some Lesser Antilles lavas subducted sediments are barely involved and enrichment in fluid-mobile elements (Ba, U, Sr, Pb, etc.) is the dominant feature. Here we evaluate whether La/Sm is a quantitative proxy of sediment involvement in volcanic arcs, and we relate dehydration and melting processes to the temperature and pressure conditions of the slab. We use Martinique as a case study because in this island both dehydration and sediment melting fingerprints coexist. We measured major and trace elements for about 130 age-constrained samples, carefully chosen to cover all volcanic phases of Martinique (25 Ma to present). Using these results we demonstrate that: (1) weathering does not modify the La/Sm ratio; (2) fractional crystallization of amphibole and/or garnet does not increase La/Sm by more than 20%; (3) rare earth element transfer from wall-rock to magma during fractionation is not significant; (4) melting of the mantle source increases La/Sm by only about 20%. As a consequence, we show that the proportion of slab sediment incorporated in the mantle wedge controls the La/Sm ratio of the source. The observed correlations between La/Sm and Nd and Hf isotopic compositions indicate that the effect of sediment addition is the overwhelming factor: La/Sm is a good proxy for slab sediment proportion in Martinique. We observe a geographical gradient between slab dehydration and sediment melting on the island. Whereas lavas located on the western side of the island display a clear sedimentary input in their source, lavas located on the eastern side of the island, closer to the trench, are clearly influenced by dehydration of the subducted slab. In addition, the aqueous fluids clearly come from the subducted basalt and they did not interact with the overlying sediments. The influence of sediment added to the source of the magmas increases from the eastern part to the western part of the island. We relate this geographical change to the pressure and temperature conditions at the slab surface. Sediments probably cross their solidus under Martinique and hydrous melting is triggered. Finally, we show that under all volcanic arcs where the signature of sediments overwhelms the signature of fluids, the slab surface reaches P-T conditions that allow the subducted sediments to melt. Inversely, under most volcanic arcs where the signal of aqueous fluids dominates over sediment melts, the subducted slab is not hot enough for the sedimentary pile to melt.

No abstract

S U M M A R YWe model using analogue experiments the subduction of buoyant ridges and plateaus to study their effect on slab dynamics. Experiments show that simple local (1-D) isostatic considerations are not appropriate to predict slab behaviour during the subduction of a buoyant ridge perpendicular to the trench, because the rigidity of the plate forces the ridge to subduct with the dense oceanic lithosphere. Oceanic ridges parallel to the trench have a stronger effect on the process of subduction because they simultaneously affect a longer trench segment. Large buoyant slab segments sink more slowly into the asthenosphere, and their subduction result in a diminution of the velocity of subduction of the plate. We observe a steeping of the slab below those buoyant anomalies, resulting in smaller radius of curvature of the slab that augments the energy dissipated in folding the plate and further diminishes the velocity of subduction. When the 3-D geometry of a buoyant plateau is modelled, the dip of the slab above the plateau decreases, as a result of the larger velocity of subduction of the dense 'normal' oceanic plate on both sides of the plateau. Such a perturbation of the dip of the slab maintains long time after the plateau has been entirely incorporated into the subduction zone. We compare experiments with the present-day subduction zone below South America. Experiments suggest that a modest ridge perpendicular to the trench such as the present-day Juan Fernandez ridge is not buoyant enough to modify the slab geometry. Already subducted buoyant anomalies within the oceanic plate, in contrast, may be responsible for some aspects of the present-day geometry of the Nazca slab at depth.