Heavy Lifting: Leveraging Machine Learning to Measure the Masses of Supermassive Black Holes

David Chemaly ( Université de Montréal )

Despite recent advances in the study of supermassive black holes (SMBH), most notably those by the Event Horizon Telescope (EHT) team, a fast and effective methodology to determine the masses of these leviathans at high redshifts continues to elude the astronomical community. Nowadays, the best method to conduct such calculations is to resolve the kinematic of the molecular gas in the region where the SMBH’s gravitational potential dominates over the galaxy’s potential. Considering how negligible the mass of a SMBH (∼10^8M☉) is compared to a host galaxy (∼10^12M☉), a high spatial resolution is required to resolve such regions which are of the order of a few tens of parsecs. This need for high-resolution data prevents us from adequately measuring masses at further distances. Here, we present a new machine learning-based method to lens observations and to resolve the surrounding gas at redshifts that go far beyond what is currently achievable. Our initial findings show that using gravitational lensing on realistic simulations provided by MassiveFIRE leads to a spatially magnified image of the targeted region. By training our new neural network on these simulated datasets, we obtained an algorithm capable of rapidly and accurately deconvolving a lensed galaxy harboring a SMBH and measuring its mass. Additionally, we treated the simulated galaxies as if they were directly observed with ALMA to enable an easy use of our model on real data. We will also discuss the implications of such a tool and showcase the surprising extent to which this new methodology can enrich our knowledge on the primary state of our universe.