Fig. 6: Sensitivity tests for the first-type model (Fig. 1).

a The time needed for the subduction zone initiation (SZI) to occur. The SZI is identified when the left-most slab surface (i.e., the reference point in Fig. 3a) is deeper than the upper plate’s bottom. b The time when slab tearing happens (e.g., Fig. 10). nan demonstrates that slab tearing does not happen in the test. c The total time needed for the subduction of the initial oceanic plate (Fig. 3a). Because subduction re-initiation does not easily happen in this scenario (Supplementary Movie 2), nan denotes that the initial oceanic plate’s subduction does not finish within 100 Myr. d Average subduction rate. The average subduction rate is calculated only in models where the initial oceanic plate completely subducts. Rx is the xth model, and the number after “−“ denotes the corresponding time. \({{{{{{\rm{\rho }}}}}}}_{{{{{{\rm{AM}}}}}}}\) is the compositional density of the asthenospheric mantle. \({{{{{{\rm{Age}}}}}}}_{{{{{{\rm{slab}}}}}}}\) is the thermal age of the initial oceanic plate⁶⁹. \({{{{{{\rm{\sigma }}}}}}}_{{{{{{\rm{slab}}}}}}}\) is the maximum yielding stress for the subducted slab (also see Supplementary Table S1). \({{{{{{\rm{\eta }}}}}}}_{\min }\) is the minimum viscosity (in Pa∙s) for the region behind the initial oceanic plate, e.g., a mid-oceanic ridge or newly formed oceanic lithosphere. In this figure, \({{{{{{\rm{\eta }}}}}}}_{\min }\) is in the form of lg (\({{{{{{\rm{\eta }}}}}}}_{\min }\)).