Résumé
Back-arc rifting is observed in extensional subduction settings, where extensional stresses have caused rupture of the overriding lithosphere. In several narrow subduction zones with a long subduction history, such as the Scotia Arc, central Mediterranean or Marianas, several ridges have been active in the course of history. Nearly instantaneously, the ridges have been jumping closer to the trench in a regular pattern. The dynamics behind this process remain unknown. We here present a new mechanism for ridge jumps, validate its feasibility in simple numerical models and finally highlight how they fit observations. We run 3D-models to simulate a long narrow slab subducting between two continental plates which retreats and creates necessary STEP-faults self-consistently. After the creation of a back-arc basin, transform faults between trench and back-arc basin form. Our results suggest that ridge jumps are a consequence of the fact that these transform faults, which link the ridge with the trench (and decoupling the overriding plate from neighboring plates), fail to remain active once the trench is too distant from the ridge. Without active transform faults, the overriding plate is coupled strongly to neighboring plates, the link between ridge and trench disappears, and a new ridge opens due to stress localization closer to the trench. In a parameter study, we show that the timing between ridge jumps are reduced for narrow slabs, and subduction fails completely for slabs narrower than 400km, where the energy to create STEP-faults is insufficient. Finally, the new ridge tends to form closer to the trench for narrower slabs, consistent with the location of maximum strain induced by the toroidal flow around the slab.