Köln 2025 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
HK: Fachverband Physik der Hadronen und Kerne
HK 20: Instrumentation V
HK 20.2: Vortrag
Dienstag, 11. März 2025, 16:00–16:15, SR Exp1A Chemie
Neural Network Corrections for Particle Identification at the ALICE TPC — •Jonathan Witte — University of Heidelberg, Germany — for the ALICE german Collaboration
The Time Projection Chamber (TPC) of ALICE (A Large Ion Collider Experiment) has been a cornerstone of the detector, providing critical insights into the physics of heavy-ion collisions for over a decade. The TPC, a gaseous detector filled with a Neon-CO2-N2 (90-10-5) mixture, offers exceptional performance and accounts for more than 90% of the raw data produced by the ALICE detector. Its capability to measure the tracks of hundreds of particles produced in each Pb*Pb collision makes it indispensable. Due to the measurement of the specific energy loss, the TPC allows for extensive particle identification (PID). The Bethe-Bloch formula accurately describes the energy loss of particles in a medium, leveraging the well-characterized properties of the TPC gas to provide up to 152 dE/dx samples per particle. This PID feature is unique among LHC detectors. To identify particles based on their dE/dx samples, a Bethe-Bloch (BB) fit is applied. However, the measured signal and precision alter strongly with environmental conditions, fluctuations in gain calibration, detector occupancy etc. Neural networks (NNs) are the technology of choice for applying corrections here. With large training datasets with clean samples of electrons, kaons, pions and protons, NNs can learn to identify patterns and anomalies, improving the calibration of PID. This talk will present the mechanism of PID of the ALICE TPC and the neural network-based corrections applied to LHC Run 3 data.
Keywords: ALICE; Time Projection Chamber; Particle Identification; Neural Network; Bethe Bloch