flares

The search for astrophysical neutrino sources is a long marathon, that started over a decade ago. With IceCube, we were already once able to find a high-energy neutrino alert in coincidence with a flaring blazar. In a search through archival IceCube data from the same direction, we then actually found a neutrino flare a few years prior to the high energy alert. The logical question to ask was then whether any other IceCube alert was revealing the same pattern with a neutrino flare from its direction.

Because the pointing of alters is not very precise, and the search space is huge, we first had to develop new statistical methods to render such an analysis possible. By employing a technique known as Expectation Maximization (ML) from unsupervised machine learning, we were able to significantly reduce the analysis complexity and able to run a fully-fledged analysis for all IceCube alert positions.

Further Information

Collaborators: Martina Karl (TUM)

Proceeding: https://arxiv.org/abs/2107.08853

Previous work: https://www.science.org/doi/10.1126/science.aat1378, https://www.science.org/doi/10.1126/science.aat2890