SEDONA-GesaRaT: A Deep Learning Accelerated Radiative Transfer Program for 3D Supernova Simulations

We present SEDONA-GesaRaT, a rapid code for supernova radiative transfer simulation developed based on the Monte Carlo radiative transfer code SEDONA. We use an atomic physics neural network, which has a convolutional neural network structure, to solve for the nonlocal thermodynamic equilibrium (NLTE) atomic physics level population calculation. SEDONA-GesaRaT is trained and validated on 119 1D type Ia supernova (SN Ia) radiative transfer simulation results. It has been applied to the 3D SN Ia explosion model N100 to perform a 3D NLTE radiative transfer calculation. The spatially resolved linear polarization data cubes of the N100 model are successfully retrieved with a high signal-to-noise ratio using the integral-based technique. The 3D LTE and NLTE radiative transfer calculations have been validated by comparisons to SEDONA program simulation results. The typical computation cost of a new 3D NLTE spectropolarimetry simulation using SEDONA-GesaRaT is only ∼3000 core-hours, which will be a factor of ∼100 faster relative to a SEDONA simulation. The large computational speed-up factor makes SEDONA-GesaRaT promising for future large-scale simulations that systematically study the internal structures of SNe. However, the ultimate advantages of SEDONA-GesaRaT will depend on the required computational cost of retraining to accommodate improvements in physical realism in radiative transfer and a wider range of SN models.