Temporal and spatial variations in crustal accretion along the Mid-Atlantic Ridge (29°-31°30′N) over the last 10 m.y.: Implications from a three-dimensional gravity study

International audienceWe have conducted a three-dimensional gravity study of the Mid-Atlantic Ridge near the Atlantis Transform to study the evolution of accretionary processes at this slow-spreading center over the last 10 m.y. We have removed from the free-air gravity anomaly the gravity contribution of the density contrast at the seafloor and the gravity contribution of the lateral density variations associated with the cooling of the lithosphere. The resulting residual gravity anomaly exhibits substantial variation along and across the ridge axis. The residual gravity anomaly can be accounted for by variations in crustal thickness of up to 3 km. For the first two segments south of the Atlantis Transform, the midportions of the segments have been associated with thick crust and the segment discontinuities have been associated with thin crust for the last 10 m.y., suggesting the segment discontinuities act as long-term boundaries in the delivery of melt to the individual segments. In contrast, our calculations indicate that for the segments north of the fracture zone, thick crust is associated with the midportions of segments and thin crust is associated with segment discontinuities only in crust less than-3 m.y. This result suggests that focused mantle upwelling has only recently developed north of the fracture zone. The onset of focused mantle upwelling at approximately 2-3 m.y. may be related to a change in the spreading direction which occurred between magnetic anomalies 5 and 3 (Figure 1) and resulted in changes in the geometry of the plate boundary north of the fracture zone. Cross sections of crustal thickness extracted along the midpoint traces of paleosegments show that, for a few segments, up to 2 km of gradual crustal thinning is observed. We suggest that the "apparent" crustal thinning is a result of lateral changes in mantle density associated with buoyant upwelling not predicted by the passive flow model used in our study. Variations in computed crustal thickness are observed across axis in all of the paleosegments in our study area, but are not correlated between individual segments. If these computed crustal thickness variations are due to temporal variations in melt production, this implies that there is little interdependence in the amount of melt supplied to adjacent segments