Institute for Advanced Study/Princeton University Early Universe/Cosmology Lunch Discussion

Virialized clusters and super-clusters of galaxies, which reside in massive dark matter halos, are the most rare objects in our Universe but contain most of its matter. As such, obtaining accurate measures of large cluster masses will put tight constraints on Lambda-CDM. A robust method for measuring the shape and orientations of massive clusters is needed in order to address systematics that complicate the interpretation of LSS measurements, such as in stacked weak lensing analysis. Furthermore, the literature has shown that shapes of dark matter halos provide a powerful cosmological probe. I will present ongoing work utilizing the MillenniumTNG simulation and machine learning to predict 3D shapes of massive clusters from cluster observables. Promising results show that a hybrid neural network is able to reconstruct 3D ellipsoidal shapes of massive clusters from 2D information.