Estimating the triaxiality of massive clusters from 2D observables in MillenniumTNG with machine leaning

Ana Maria Delgado 
Post Doctoral Fellow

Abstract: Properties of massive galaxy clusters, such as mass abundance and concentration, are sensitive to cosmology, making cluster statistics a powerful tool for tightening constraints on cosmological parameters. However, the standard approach of favoring a spherically symmetric shape with which to model galaxy clusters can lead to biases in the estimates of cluster properties. In this work, we present a machine-learning approach for estimating the 3D triaxiality of massive galaxy clusters from 2D observables. We utilize the flagship volume of the cosmological-hydrodynamical MillenniumTNG (MTNG) simulation suite as our ground truth. Our model combines the feature extracting power of a convolutional neural network (CNN) and the message passing power of a graph neural network (GNN) in a multi-modal fusion network. Our network achieves triaxial shape and orientation estimates which model MTNG clusters at a ~30% improvement compared to spherically symmetric approximations. Our results provide optimistic outcomes for improved measurements of clusters which populate the high end of the halo-mass function.

Host: Ankita Bera