Data-driven magnetohydrodynamic simulation of the initiation of a coronal mass ejection with multiple stages

Coronal mass ejections (CMEs) are the primary drivers of adverse space-weather events, yet their initiation and onset prediction remain insufficiently understood due to the complexity of the magnetic topology and physical processes in real solar source regions. Using a fully observational-data-driven magnetohydrodynamic simulation, we successfully reproduced the initiation of a CME originating from the super active region 13663, with only a one-minute time lag between the flare peak in observations and the velocity peak of the rising flux rope in the simulation. Moreover, the eruptive structure exhibits a multi-stage kinematic evolution: an initial slow acceleration, a plateau at a nearly stationary height, and a subsequent impulsive acceleration. These stages correspond to torus instability, the downward tension force exerted by the overlying toroidal field, and fast magnetic reconnection, respectively. Our results highlight the inherently multistage nature of CME initiation in real events. In configurations with strong overlying toroidal fields, the downward toroidal-field-induced tension force can suppress the rise of the flux rope and produce a plateau phase at a nearly stable height, even when torus instability occurs. In contrast, the subsequent fast magnetic reconnection beneath the flux rope can drive the impulsive eruption more effectively. The close agreement between the observed and simulated peak times over one minute demonstrates the strong potential of our data-driven model for predicting CME onset.