Speaker
Description
Galaxy morphology reveals insights into a galaxy’s properties and evolutionary history, however, galaxies (and their morphologies) evolve on timescales much longer than human lifetimes. Therefore studying the evolution of individual galaxies is difficult as we only see one brief stage of the galaxy’s evolution. We are fortunate that we are capable of observing many galaxies across cosmic time, all showing a small slither of the evolutionary process, we use this fact to trace galaxy evolution. Using an unsupervised autoencoder trained on galaxy images from the IllustrisTNG-50 simulation, we select galaxies at low redshift and then perform a repeating search where we use the learned representation space of our selected galaxies to find visually and morphologically similar galaxies at a slightly higher redshift, these new galaxies are used to find similar galaxies at a slightly higher redshift etc. The repeated search creates a chain of galaxies that spans a large redshift range, creating a plausible evolutionary pathway for the initial low-redshift galaxies. This method allows us to take low-redshift galaxies and find possible pseudo-progenitors at high redshift, allowing us to measure how the galaxies might have evolved based on their current morphology. Applying this method to real survey data could uncover the possible evolutionary pathways of galaxies.