Title: Artificial Intelligence Assisted Inversion (AIAI) of Synthetic Type Ia Supernova Spectra

Abstract: We generate $\sim$ 100,000 model spectra of Type Ia Supernovae (SNIa) to form a spectral library for the purpose of building an Artificial Intelligence Assisted Inversion (AIAI) algorithm for theoretical models. As a first attempt, we restrict our studies to time around $B$-band maximum and compute theoretical spectra with a broad spectral wavelength coverage from 2000 $-$ 10000 ${\rm \AA}$ using the code TARDIS. Based on the library of theoretically calculated spectra, we construct the AIAI algorithm with a Multi-Residual Convolutional Neural Network (MRNN) to retrieve the contributions of different ionic species to the heavily blended spectral profiles of the theoretical spectra. The AIAI is found to be very powerful in distinguishing spectral patterns due to coupled atomic transitions and has the capacity of quantitatively measuring the contributions from different ionic species. By applying the AIAI algorithm to a set of well observed SNIa spectra, we demonstrate that the model can yield powerful constraints on the chemical structures of these SNIa. Using the chemical structures deduced from AIAI, we successfully reconstructed the observed data, thus confirming the validity of the method. We show that the light curve decline rate of SNIa is correlated with the amount of $^{56}$Ni above the photosphere in the ejecta. We detect a clear decrease of $^{56}$Ni mass with time that can be attributed to its radioactive decay. Our Code and model spectra are made available on the website \href{https://github.com/GeronimoChen/AIAI-Supernova}{this https URL}.