Towards automating abundance tomography for type Ia supernovae

In this thesis we present techniques to automate the tting of spectral time series of
Type Ia supernovae (SNe Ia). These transient objects play an important part in the
chemical evolution of the Universe and are used as important distance indicators to map
out the expansion history of the Universe. Despite their importance, many aspects of
these transient events are still unknown. An important tool to study the explosion in
detail is the abundance tomography method in which a model is tted to the spectral
time series of an observation using a fast spectral synthesis code.
In this work we present rst steps to accelerate and automate this method in an attempt
to explore the parameter space surrounding the best tting set of parameters and to
highlight degeneracies. To this end, we develop a framework that uses Machine Learning
(ML) techniques to quickly generate spectra and then use Markov chain Monte Carlo
(mcmc) methods to explore the parameter space with the help of Bayesian statistics to
construct a measure of the quality of t. After extensively testing and validating the
framework, we apply it to supernova (SN) 2002bo, which was already the subject of
similar studies done manually. However, reproducing the results found in literature is a
challenge that needs further research beyond this thesis.