To address the question of the initiation and mechanisms involved in the process of subduction zone formation, we explored most of the available evidence for the subduction initiation (SI) during the Cenozoic. For this, we targeted a total of 70 candidate sites for subduction initiation cumulating ~70,000 km of trench, two thirds of which are still active and a majority still immature. Our strategy is to define four stages reached for each subduction initiation site (SIS) from the incipient-diffuse stage through incipient-localized stage and early arc magma production to self-sustained subduction. We have paid special attention to prematurely extinguished, i.e., aborted, subduction attempts in order to better understand the reasons for the termination of the process, and thus to clarify the conditions of success. The failure of SI results from a combination of hindering parameters (e.g., lithosphere cooling, frictional resistance, unfavorable age contrasts for intra-oceanic SISs) and insufficient external forcings (e.g., too low convergence velocity). From this comprehensive study, we find that new subduction zones regularly nucleate, at a mean rate of about once every Myr, and with a success rate of more than 70% to reach subduction maturity, generally in less than ~15 Myr, ~3-8 Myr for the shortest time between the very early stage and the self-sustained stage. A majority forms at the transition between an ocean and a continent, plateau or volcanic arc, demonstrating that large differences in composition, topography and/or lithospheric weaknesses favor the localization of the strain. Lithospheric forces are required to ensure the success of the process in the early (immature) stages, with the help of mantle forces in a third of the cases. Multiple triggers are common. Stress during the SI process is compressive in most, if not all, cases and oriented obliquely to the nascent plate boundary in more than half of the cases. The incipient plate boundary generally reactivates an old lithospheric fault, most often with a change in its kinematics, i.e., conversion of a transform plate boundary, a former normal or a detachment fault, or even a former spreading center. Sometimes, the new lithospheric fault reactivates a former subduction fault. There is no rule concerning the age of the subducting plate which varies from 0 to 140 Ma in the examples studied. In the same vein, the subducting plate is not necessarily older than the overriding plate when it is oceanic. Both situations are equally observed.

为了解决俯冲带形成过程中的起始和机制问题，我们探索了新生代俯冲起始（SI）的大部分可用证据。为此，我们针对总计 70,000 公里长的海沟进行了总共 70 个俯冲起始候选地点，其中三分之二仍然活跃，大多数仍然不成熟。我们的策略是定义每个俯冲起始位点 (SIS) 所达到的四个阶段，从初始扩散阶段到初始定位阶段和早期弧形岩浆产生再到自我持续俯冲。我们特别关注过早熄灭，即中止的俯冲尝试，以更好地了解过程终止的原因，从而阐明成功的条件。SI 的失败是由阻碍参数（例如，岩石圈冷却、摩擦阻力、海洋内 SIS 的不利年龄对比）和外强迫不足（例如，收敛速度太低）的组合造成的。从这项综合研究中，我们发现新的俯冲带有规律地成核，平均速度约为每 Myr 一次，并且达到俯冲成熟度的成功率超过 70%，通常小于 ~15 Myr，~3- 8 Myr 表示极早期阶段和自我维持阶段之间的最短时间。大多数形成于海洋和大陆、高原或火山弧之间的过渡处，表明成分、地形和/或岩石圈弱点的巨大差异有利于应变的定位。需要岩石圈力来确保早期（未成熟）阶段的过程成功，在三分之一的情况下借助地幔力。多个触发器很常见。SI 过程中的应力在大多数（如果不是全部）情况下是压缩的，并且在超过一半的情况下倾斜于新生板块边界。初始板块边界通常会重新激活旧的岩石圈断层，最常见的是其运动学发生变化，即转换板块边界、以前的正常断层或分离断层，甚至以前的扩张中心。有时，新的岩石圈断层会重新激活以前的俯冲断层。在所研究的例子中，没有关于俯冲板块年龄从 0 到 140 Ma 变化的规则。同样，俯冲板块在海洋时不一定比上覆板块更老。这两种情况同样被观察到。