Application of Principal Component Analysis to Galaxy Spectral Energy Distributions

Galaxy spectra are a useful diagnostic tool that can be used to reveal the intrinsic properties of galaxies, such as their star formation rate and stellar mass, along with the conditions in the interstellar medium. Generally the computation of the full galaxy spectra within galaxy formation and evolution models tends to be very time consuming and memory inefficient, so the cal- culation of spectra is typically only done in post-processing for a subset of model galaxies (e.g. Trayford et al. 2017, Cowley et al. 2018). Upcoming sur- veys will measure tens of millions of spectra (e.g., Euclid (Laureijs et al. 2011) and the Dark Energy spectroscopic Instrument (DESI; Levi et al. 2019)). To exploit these data, theoretical models need to be able to predict spectra to connect more closely with these surveys. In this thesis, we aim to reduce the computational expense when calculating galaxy spectra by applying principal component analysis (PCA) to the spectral energy distributions of simple stel- lar populations (SSPs). We consider different star formation histories and different matallicities. As a result, we find that the dimensionality of the SSP spectra can be reduced by a factor of ∼50 whilst there is only a small loss in accuracy (∼1 - 5%) of the reconstructed spectra. Moreover, we find that this loss in accuracy is negligible when computing broadband magnitudes (≪ 1%). Our results suggest that this calculation method may be a plausible way to predict spectra for all the galaxies in the output of a semi-analytical model covering a cosmological volume (e.g. GALFORM; Cole et al. 2000).