Deep Learning on Type Ia Supernovae

The thesis presents the design of an Artificial Intelligence Assisted Inversion (AIAI) method to estimate type Ia supernova (SN Ia) ejecta structure based on the observed optical spectral time sequence. The research applied neural networks to 126 SNe Ia and found a correlation between the 3700 A spectral feature and the 56Ni elemental abundance. To further adapt the AIAI method to the SNe Ia 3D structure estimate, the author developed an integral-based technique to significantly increase the signal-to-noise ratio in the polarized time-dependent 3D radiative transfer computations. To understand the SNe Ia progenitors, the spatially resolved SN Ia host galaxy spectra from MUSE and MaNGA were employed to estimate the delay time distribution (DTD). By using a grouping algorithm based on k-means and earth mover's distances, the research separated the host galaxy stellar population age distributions into spatially distinct regions and used the maximum likelihood method to constrain the DTD. It was found that the DTD is consistent to the double-degenerate progenitor models.