Description
Ultra-high-intensity multi-petawatt lasers (I ≥ 10^22 W/cm2) enable access to the relativistic self-induced transparency regime. In this regime, initially opaque plasmas become transparent as electrons gain relativistic effective mass, lowering the plasma frequency. This fundamentally alters electron dynamics and allows the laser to propagate through increasingly dense targets. We employ 2D particle-in-cell simulations to systematically investigate laser energy coupling to electrons across varied laser and target parameters. To circumvent the prohibitive computational expense of a high-resolution grid scan, we efficiently explore the parameter space using random sampling and build a surrogate model via Gaussian process regression coupled with active learning.