Speaker
Description
The PIONEER experiment is a next-generation investigation into rare pion decays, aimed at probing beyond the Standard Model (SM) physics by accurately measuring the charged-pion branching ratio of electrons versus muons Re/$\mu$ - providing a sensitive test of Lepton Flavour Universality. The SM provides a calculation of Re/$\mu$ at the 0.01% level, a factor 15 times better than the most precise experimental measurement. The PIONEER experiment aims at closing the precision gap between theory and experiment, with the possibility of revealing new physics at the PeV scale. PIONEER has been proposed and approved with high priority at the Paul Scherrer Institute in Switzerland.
This poster will focus on applying machine learning (ML) techniques to PIONEER’s active target (ATAR). The ATAR is a central piece of the detector and a key enhancement over earlier experiments measuring Re/$\mu$. It is composed of 48 successive planes of 120$\mu$m thin low-gain avalanche diodes (LGADs). This technology allows 5-dimensional event reconstruction providing precise timing, energy, and three-dimensional spatial information. ML offers an attractive alternative to traditional reconstruction and analysis approaches for pattern recognition. This work explores the integration of a neural-network architecture based on transformers into the ATAR reconstruction pipeline, and investigates its performance and potential for particle identification and tracking.
| Keyword-1 | Machine Learning |
|---|---|
| Keyword-2 | Active Target |