Conditions for subduction zone initiation in present-day Earth in the light of Cenozoic examples and numerical simulations

The process of subduction initiation is still debated and caused a great deal of controversy such as: Can a subduction zone initiate without any external forcing? Is the thicker and more buoyant lithosphere really the more likely to subduct? To try to answer these questions, a database of 70 cases of Cenozoic subduction zone initiation was built and analyzed in 2021. We find that initiation of subduction zone succeeded in reaching the mature stage for 72% of the cases, usually in less than ∼15 Myr, and that compositional heterogeneities are essential to localize convergence. Interestingly, we observe that the plate age offset when convergence starts is very low (close to zero) in half the cases; otherwise the incipient downgoing plate is as often the younger lithosphere as the older one, and that it could have any age. This indicates that the buoyancy contrast does not determine the subduction zone polarity.

 

We then build a numerical experimental setup to try to explain this observation. We consider the simple set-up of an oceanic transform fault (TF)  or a fracture zone and perform 2D thermomechanical simulation, by combining a non-Newtonian ductile and a pseudo-brittle rheologies. We carry out three different and complementary studies.

We first study the feasibility of ’spontaneous’ subduction initiation, i.e., gravitational collapse of the older lithosphere, at a TF. Simulations show that the main mechanical parameters have to be tuned to quite extreme values to trigger the old lithosphere collapse. The comparison to the geological records of the 3 most likely candidates of ’spontaneous’ subduction initiation (Izu-Bonin-Mariana, Yap, and Matthew & Hunter) leads us to conclude that this mode of initiation at a TF is unlikely in modern Earth conditions.

 

Second, we simulate normal convergence symmetrically imposed on the two oceanic plates forming the TF to study the ’forced’ mode of subduction initiation.  Surprisingly, the range of conditions leading to the older plate subduction is quite limited, whereas the subduction of the younger plate is much more frequently simulated. We find that the success of initiation, as well as the subduction zone polarity strongly depends on the plate age pair and on the initial structure of the TF. The rheological properties and the plate ability to be deformed and sheared may be the first order parameters controlling the subduction initiation mode for rather stiff lithospheres. The model predictions are in good agreement with different Cenozoic records (Gagua, Mussau, and Hjort).

However, this modeling forecasts that the older plate subduction cannot occur at a typical TF for large plate age offsets, in disagreement with what is observed at Izu-Bonin-Mariana, Matthew & Hunter or Palau. We note that, in these cases, a thicker crust made of continental or oceanic terranes (fossil arc or plateau) was always present near the inter-plate domain when convergence started. The third numerical study investigates how such a lithologic ‘raft’ might affect subduction initiation for high plate age contrasts. We find that the raft dimensions and location basically control the under-thrusting of the older and thicker plate.