Mining The Spectrum

: From Stellar Parameters to the Physical Properties of Galaxies

Abstract

This thesis provides a comprehensive analysis of both stellar and galactic spectra, which are crucial for probing galaxy formation and evolution processes in the Universe. Using the Mapping Nearby Galaxies at Apache Point Observatory Stellar library (MaStar), the most extensive stellar library to date, we de-termine the physical properties (Effective Temperature, Surface Gravity and Chemical Composition) of 59,266 stellar spectra belonging to 27,945 unique stars. We demonstrate that extracting these proper-ties through full spectral fitting of model atmospheres to observed spectra is challenged by systematic offsets and parameter degeneracies. One notable degeneracy, which poses a significant obstacle across the optical wavelength range of MaStar (3622-10354Å), is the dwarf-giant degeneracy. We perform an in-depth analysis of various methodologies, from Markov Chain Monte Carlo to a simple 𝜒 2 minimisation, to overcome these issues. By incorporating astrophysical priors, such as PARSEC isochrones, and target-ing specific spectral features like the Na I doublet, we effectively break this degeneracy. This results in a reliable catalogue of stellar atmospheric parameters. These parameters form the foundation of new Stel-lar Population Models, which we then rigorously calibrated against globular clusters. Using the Fitting IteRativEly For Likelihood analYsis (FIREFLY) full spectral fitting code, we test the ability of the models to recover the ages and chemical compositions of the clusters. Additionally, we conduct a novel compar-ison between FIREFLY and two other widely-used spectral fitting codes, STARLIGHT and pPXF. We find that FIREFLY is considerably less prone to over-fitting than the alternative codes, and that a lower 𝜒 2 does not necessarily equate to a more accurate fit. Finally, FIREFLY has been computationally upgraded with a highly parallelised architecture to handle Dark Energy Spectroscopic Instrument (DESI) datasets. We processed approximately 3 million galaxies from DESI’s Data Release 1, and we initiated the creation of a Value-Added Catalogue. Contributing to a DESI Key Paper, we recovered the physical parameters of galaxies across multiple tracers. This analysis involved conducting cosmological consistency checks, which demonstrated that FIREFLY recovers physically plausible ages. This contrasted with other codes, which often returned ages exceeding the age of the Universe. Ultimately, this thesis demonstrates that full spectral fitting, constrained by astrophysical priors, remains the best method for decoding complex stellar populations.

Date of Award	26 Jun 2026
Original language	English
Awarding Institution	University of Portsmouth
Supervisor	Claudia Maraston (Supervisor), Daniel Thomas (Supervisor) & Adam Amara (Supervisor)