Résumé
Olivine-rich troctolites (Ol > 70%, Ol T) indicate that extensive melt impregnation of preexisting Ol rich lithologies participate to the building of slow spread crust. To constrain their origin and their impact on the structure and geochemistry of oceanic crust, we realized a multi-scale petro-structural, geochemical, and numerical modelling study of Ol T drilled at IODP Hole U1309D (Atlantis Massif, Exp. 304/305). Ol T display deformed (high temperature imprint) corroded coarse grained to undeformed fine grained Ol embayed in poikilitic Cpx and Plg. Ol crystallographic preferred orientations show [001] cluster suggesting formation after impregnation and assimilation of a deformed Ol rich matrix at high melt/rock ratios. Ol have variable major and minor element compositions, but similar fractionated REE (Dy (sub N) /Yb (sub N) = 0.04 - 0.11). Chemical traverses along principal crystallographic axes of Ol are flat suggesting local equilibrium between Ol and neighboring phases. 3 types of Ol T were distinguished. Ol T 1 - 2 display sharp contacts. Ol T 1 has Ol <75% (single grains) and primitive compositions (Mg# = 85-86; Ni = 1870-2840 ppm, Mn = 1570-1950 ppm; Li = 1.2 - 2.7 ppm). Ol T 2 have high Ol ( > 75%, dominantly aggregates) yet more evolved composition (Mg# = 83-84, Ni = 1790 - 2510 ppm, Mn = 1760 - 1990 ppm, Li = 1.5 - 3.9 ppm) in contrast to modal and composition trends predicted by MORB crystallization. Ol T 3 has diffusive contacts with gabbroic veins, variable modal Ol with the most evolved compositions and record late stages of Ol-T formation. Ol T compositions are best modelled assuming percolation of primitive MORB melts into Hole U1309B harzburgite, triggering Opx dissolution, followed by Ol assimilation and Plg + Cpx crystallization. Modelling shows that Ol Ni variations at constant Mg# in Ol T are mantle inherited. Ol T 1 compositions were fitted assuming higher Ol assimilation (M (sub a) = 0.06 - 0.13) in contrast to Ol T 2 -3 (M (sub a) = 0.01 - 0.02). Ol T 3 was "buffered" by more evolved melts and shows cooling of reacted melts. We interpret the 3 Ol T types as resulting from initial local spatial variations in mantle permeability (pyroxene distribution?) which in turn controlled melt transport and mantle-melt interactions. We expect that such reactive percolation processes will shift MORB compositions to apparent high pressure fractionation.