Galaxy merging is a fundamentally important physical process of galaxy formation and evolution. The signature of such merging can be imprinted on the physical properties of outer parts of galaxies (e.g., stellar halos and globular cluster systems around galaxies). However, it is a formidable task for astronomers to detect such faint signals of ancient merger events of galaxies.
In this project, students will try to “hunt for’’ fossil records of ancient galaxy merging in stellar structures and kinematics of galaxies by applying deep convolutional neural networks (CNNs) to the 2D images of galaxies. First, students will create a large number of 2D images of galaxies for isolated and merging galaxies for training CNNs based on the results of computer simulations. Using new data sets from simulations, students then will investigate whether a (new) simulated galaxy can be correctly classified as an isolated galaxy or a merging galaxy. The main purpose of this project is to develop a CNN for such classification.