Dissertation: Statistical Methods in Quantitative Spectroscopy of Solar-Type Stars

  • Date:
  • Location: Zoom: https://uu-se.zoom.us/my/alvinzoomroom
  • Doctoral student: Alvin Gavel
  • Organiser: Institutionen för fysik och astronomi
  • Contact person: Alvin Gavel
  • Disputation

Dissertation by Alvin Gavel.

Galactic archaeology is the research field that attempts to reconstruct the history of the Milky Way, using primarily the tools of astrometric studies and chemical studies. The latter in turn uses stellar spectroscopy. Thanks to technological advances, the field of stellar spectroscopy now has access to much larger amounts of observational data than it used to. At the same time, also thanks to technological advances, the field able to use increasingly more sophisticated modelling. This opens up for the possibility of attacking research problems in Galactic archaeology that were previously intractable. However, it also creates a problem: Access to greater amounts of data means that the random errors in studies will tend to shrink, while the systematic errors tend to stay of the same size. At the same time, improvements in modelling means that studies can look for increasingly subtle effects in their data.

Each article in this thesis attempts solve some specific problem within Galactic archaeology - where possible also developing a general method for handling that type of problem in a way that takes systematic errors into account. In Article I we document a code for estimating stellar parameters from spectra observed with UVES. We use a set of benchmark stars to evaluate the performance of the pipeline, and develop a general method for benchmarking similar codes. In Article II we estimate elemental abundances in spectra in the globular cluster M30 as a means of estimating the Parameter T0 in AddMix models of stellar evolution. At the same time we develop a general method for taking into account systematic errors in derived abundances when estimating parameters in stellar evolution models. In Article III we test whether it is possible to use machine learning to estimate alpha abundances from low-resolved BP/RP spectra from the Gaia satellite.