3D full-GR simulations of magnetorotational core-collapse supernovae on GPUs: a systematic study of rotation rates and magnetic fields

ABSTRACT We present a series of fully three-dimensional high-resolution dynamical-spacetime general-relativistic magnetohydrodynamics (GRMHD) simulations of core-collapse supernovae (CCSNe) for a progenitor of zero-age main-sequence (ZAMS) mass $25\, \mathrm{M}_\odot$. We simulate a total of 12 models to systematically study the effect of rotation rates and magnetic fields on jet formation via the magnetorotational mechanism. We have performed simulations on OLCF’s Frontier using the new GPU-accelerated dynamical-spacetime GRMHD code GRaM-X for magnetic fields $B_0 = (10^{11}, 10^{12})\, \mathrm{G}$ and rotation rates $\Omega _0 = (0.14, 0.5, 1.0, 1.5, 2.0, 2.5)\, \mathrm{rad\, s}^{-1}$. We find that models with $B_0=10^{11}\, \mathrm{G}$ fail to explode, while those with $B_0=10^{12}\, \mathrm{G}$ show a wide range of jet morphologies and explosive outcomes depending on the rotation rate. Models with $B_0=10^{12}\, \mathrm{G}$, which have $\Omega _0=(0.14,0.5)\, \mathrm{rad\, s}^{-1}$, also fail to explode. Models with $B_0=10^{12}\, \mathrm{G}$ and $\Omega _0=(1.0,1.5)\, \mathrm{rad\, s}^{-1}$ form tilted jets, giving the ejecta a more spherical character, but one also different from predominantly neutrino-driven explosions. Models with $B_0=10^{12}\, \mathrm{G}$ and $\Omega _0\ge 2.0\, \mathrm{rad\, s}^{-1}$ form jets that show ejecta velocities $\gtrsim 15\,000\, \mathrm{km\, s}^{-1}$, making them suitable candidates for broad-lined Type Ic supernova progenitors. We also perform comparison lower resolution simulations. While average shock radii, protoneutron star (PNS) magnetic fields, PNS rotation rates, and the kink instability have similar behaviour as in high-resolution simulations, the ejecta mass and energy are $\rm{approximately\ two\ to\ three}$ times lower. This work represents the largest set of 3D GRMHD simulations studying magnetorotational supernovae in full GR and demonstrates the potential of systematic studies with GPU-accelerated 3D simulations of CCSNe.