Understanding real galaxy mergers through simulations and deep learning

Galaxy mergers are events that underpin how our Universe grows and evolves. Despite their importance, knowing exactly what is and is not a real galaxy merger is not as simple as it seems. Equally, knowing how far along in a merger is for real galaxy mergers is a non-trivial issue, beyond a simplistic pre-coalescence, coalescence and post-coalescence categorisation. Due to the long times scales involved, on the order of a billion years, we cannot wait and watch mergers occur and evolve to solve these issues. Thus, merger detections can be uncertain and it is difficult to follow the physical processes inside mergers as a merger progresses. Simulations do not have these problems. Zoom in simulations allow us to closely follow two colliding galaxies and observe how physical properties change and evolve. Cosmological simulations allow us to gather a large number of known galaxy mergers for statistical samples. Thus, being able to use simulations with known truths of if a galaxy is a merger and known times before or after a merger event to identify and classify galaxy mergers would be a powerful tool. This paper will discuss how we can take the images and truths from simulations and apply them to the real Universe through deep learning. It will explore how this can allow us to gain a base truth for what is a merger and identify how far along real mergers are. From this, we can have a sneak peak at the kind of science that is right around the corner using these cutting edge techniques.