Preconditioning of subduction zone initiation at passive margins by gravitational instabilities

Subduction zone initiation (SZI) represents a critical step in the evolution of plate tectonics, yet its controlling mechanisms remain debated. While SZI is commonly classified as induced or spontaneous depending on the dominance of far-field convergence or local buoyancy forces (Stern, 2004; Stern and Gerya, 2018), geological and numerical studies suggest that purely spontaneous subduction at passive margins is unlikely under present-day conditions (Arcay et al., 2020; Lallemand and Arcay, 2021). Nevertheless, passive margins are characterised by strong lateral contrasts in density, rheology, thermal structure and sedimentary loading, which may generate gravitational instabilities capable of locally weakening the lithosphere. In this study, we investigate whether gravitational instabilities at passive margins can act as a preconditioning mechanism for subduction, facilitating induced SZI and influencing the early evolution and geometry of the subduction zone once convergence is applied. We perform several hundred two-dimensional thermo-mechanical simulations using the finite-element code FALCON (Regorda et al., 2023), modelling a passive margin. The models include an initial gravitational phase, followed by an induced convergence phase with velocities ranging from 0.01 to 1 cm/yr. To systematically explore lithospheric weakening, we vary viscous weakening intervals and plastic weakening laws, allowing us to quantify deformation localization through strain-rate analysis near the margin. Our results show that, for sufficiently weak rheological configurations, gravitational instabilities lead to transient strain-rate localization within the passive margin, controlled by plastic weakening at shallow levels and viscous weakening at depth. The mechanically damaged zone may be efficiently reactivated when convergence starts. In these cases, subduction initiates and develops readily into a coherent subduction interface, particularly at moderate to high convergence rates.