To achieve our two goals of measuring the cosmic expansion rate and constraining SN progenitors via gravitationally lensed SNe, we need to find these rare astrophysical events. The chance of lensed SNe occurring is extremely low since SNe and gravitational lensing are both inherently rare events. However, current and upcoming astronomical surveys are monitoring large areas of the night sky and imaging billions of galaxies, which would capture lensed SNe. Our strategy is to find lens systems, and then wait for one of the background sources to explode. We have searched for gravitationally lensed galaxies in the Pan-STARRS and Hyper-Suprime-Cam (HSC) surveys that cover the entire northern sky. To cope with the huge data volume, we have used Deep Learning to quickly classify images of astrophysical objects. From our newly developed tools, we have found >800 new lens candidates in the Pan-STARRS and HSC surveys that could serve as potential hosts of lensed SNe (Cañameras et al. 2020; 2021; Shu et al. 2022). We have monitored these lens candidates with the Zwicky Transient Facility (ZTF), a survey that maps out the northern sky every 2-3 days, as a way to detect a SN occurring in one of the lens systems.
Since ZTF only reaches the brighter and nearer galaxies, we have not found a suitable lensed SN in ZTF, as expected. However, we expect to discover at least several lensed SNe per year from the upcoming LSST, and the lens search methods that we have developed are readily applicable to LSST. Our project has investigated the observing strategy of LSST to maximize the number of lensed SNe (Huber et al. 2019; Lochner et al. 2022). The results of these studies have been used as input for optimising the LSST observing strategy.
The time delays between the lensed SN images are needed for our two scientific aims. We have determined follow-up observation requirements for measuring time delays (Huber et al. 2019; 2020). Accounting for microlensing, we have developed novel methods to measure time delays based on machine learning that outperform existing methods for Type Ia SNe (Huber et al. 2022; Huber & Suyu, submitted) and based on spectral evolution of Type IIP SNe (Bayer et al. 2021). In addition, our extensive simulations show that the effect of microlensing is negligible at the beginning of SNe, enabling us to constrain SN progenitors (Suyu et al. 2020). We have also developed fast ways to model the mass distribution of the lens system using machine learning (Schuldt et al. 2021; 2023ab). We have further automated our modelling procedure to enhance efficiency (Ertl et al. 2023). Both developments on the machine learning and conventional modelling approaches facilitate the analyses of lensed SNe for achieving our scientific goals, as demonstrated by the application to the two newly discovered lensed SN systems: SN Zwicky (Pierel et al. 2023) and SN H0pe (Pierel et al. 2024).
For cosmological studies, we have investigated the cosmological information expected from a sample of lensed SNe in the LSST survey (Suyu et al. 2020). We have also demonstrated the combination of SNe and gravitational lensing to provide a measurement of the expansion rate of the Universe that is less sensitive to model assumptions (Taubenberger et al. 2019; Wong et al. 2020). We have developed self-consistent lensing and dynamical model software that will be crucial to break model degeneracies for cosmological inference (Wang et al., in prep.). Utilising the first strongly lensed SN, SN Refsdal, we have demonstrated the lensed SN by a galaxy cluster accurately constrains both the expansion rate and the geometry of the Universe (Grillo et al. 2024).
