Physical Society Colloquium

Learning to See the Dark Matter in Galaxy Clusters

Hy Trac

Department of Physics
Carnegie Mellon University

Galaxy clusters contain large amounts of cold dark matter, hot ionized gas, and tens to hundreds of visible galaxies. In 1933, Fritz Zwicky postulated the existence of dark matter when he inferred the total mass of the Coma cluster from the motions of its galaxies. We now think that dark matter makes up about 85% of the total matter, but we have yet to map out its spatial structure. How are the dark matter and baryons distributed within massive galaxy clusters? In this talk, we first provide an update on the mass of the Coma cluster using modern AI/ML techniques. A convolution neural network is used to train and test on the entire distribution of galaxy positions and velocities, while bayesian deep learning is used to infer the posterior likelihoods for cluster mass. Second, we show how generative diffusion models can be trained on multi-wavelength images (e.g. SZ effect, X-ray emission, gravitational lensing) of galaxy clusters to predict the gas, dark matter, and total matter projected density fields. When applied to synthetic images of simulated clusters, the inferred mass reconstructions are accurate and unbiased. Mapping the unknown in galaxy clusters with AI/ML is promising. Knowing where is the dark matter will help us to understand its nature and that of the Universe.