Abstract:

Lensed supernovae are ideal to investigate the supernova progenitor systems and for cosmological studies such as measuring the Hubble constant H0. To measure H0 with percent-level precision, the combination of multiple systems is needed, as already done with galaxy-quasar systems by the H0LICOW and TDCOSMO collaborations. While so far most lensed supernovae were detected only by chance and not through a dedicated search, dedicated effort is required for a sample that allows a combination of measurements. Since detecting these peculiar lenses through the supernova brightness often leads to small image separation systems with unresolved images that have too short time delays for measuring the Hubble constant, we present an alternative approach carried out within the HOLISMOKES collaboration. We use all detected transients to cross-match with all known static lenses on a daily basis. For this procedure, dedicated and efficient lens search projects are a crucial step. I will introduce our ongoing search for gravitationally lensed transients using deep learning, where I have combined ground-based and space-based imaging using the Hubble Space Telescope (HST) and Legacy survey data, simulating Euclid and the Vera C. Rubin Observatory (LSST). I will present the steps and the different deep learning architectures that have been tested, and also summarize efforts from the whole community.