Predicting intermediate-mass black hole formation in star clusters with machine learning

The presence of intermediate-mass black holes in nearby star clusters has been the source of a decades-long debate. We use neural network and random forest regression models trained on simulated data from the Rapster evolution code to predict the masses of the heaviest compact objects assembled via successive black-hole mergers based on observed cluster properties such as total mass and half-mass radius. We present predictions on the possible intermediate-mass black hole population in nearby globular and nuclear star clusters. Our findings suggest that globular clusters rarely harbor massive (≳ 100M⊙) BHs with an occupation fraction of about 0.02. In a few nuclear star clusters, such as NGC 5102 and NGC 5206, we find MBH ≳ 100M⊙, and we compare our predictions with kinematically-inferred BH mass estimates. If BHs as massive as previously claimed harbor these environments, we advocate for growth channels other than repeated BH mergers, such as accretion of gas and stars.