Orateur
Description
Superluminous Supernovae (SLSNe) are rare and extremely bright stellar explosions. Their precise powering mechanism remains an open question, with models suggesting contributions from magnetars, pair-instability explosions, or interaction with dense circumstellar material. Modern wide-field optical surveys, such as the Zwicky Transient Facility (ZTF), have allowed the automatic photometric detection of many SLSNe events, contributing greatly to the constraints on our models. The Vera C. Rubin Observatory (LSST) will significantly increase the number of SLSNe candidates by probing our Universe more deeply. It is expected to detect 104 SLSNe every year, which will deeply improve our knowledge on them.
In such big data surveys, efficiently identifying SLSNe among the vast number of transient alerts is very challenging and tailored pipelines should be constructed. Ideally, they should be fast enough to process large amounts of data, and interpretable enough to be trusted and used by the community.
We address this challenge by using a curated sample of spectroscopically confirmed SLSNe to develop a machine-learning classification framework for ZTF data. The model has been integrated into the Fink ecosystem, where it has already been operating in real time on ZTF alerts and has shared candidates publicly in dedicated channels. Over the past months, the classifier has demonstrated very high completeness and has assisted domain experts in identifying SLSN events. In this contribution, we present the methodology, performance, and limitations of the model, and we also discuss future prospects. In particular, its integration within the Fink environment facilitates quick and straightforward adaptation for the upcoming LSST era, which will represent a crucial period for scientific discoveries.