Köln 2025 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

HK: Fachverband Physik der Hadronen und Kerne

HK 16: Heavy-Ion Collisions and QCD Phases III

HK 16.3: Talk

Tuesday, March 11, 2025, 14:30–14:45, HS 3 Chemie

Dielectron Identification with Machine Learning in Ag+Ag collisions at 1.58A GeV at HADES — •Henrik Flörsheimer for the HADES collaboration — Technische Universität Darmstadt

The High-Acceptance-Di-Electron-Spectromet (HADES) is a fix target experiment capable of measuring heavy-ion as well as elementary collisions. With beam energies around a few GeV nuclear matter at high densities and moderate temperature can be observed. One way to gain information about these collisions is to study electro- magnetic probes, such as the virtual photon decaying into electron positron pairs. They can be used to characterize the evolution of the fireball or to gain further information using their invariant mass spectrum to determine a fireball temperature or potential in medium modifications.

At HADES, the main components for reconstruction of dileptons are the ring imaging Cherenkov (RICH) detector, two Multi-wire drift chambers (MDCs) before and after the magnet for tracking and momentum determination, an electromagnetic calorimeter (ECAL) measuring the energy loss, and a forward wall for determining the centrality of the collisions.

In this contribution, we discuss new methods to utilize all these detector observables in a multivariate analysis in order to optimally identify leptons. We demonstrate how the performance in the dilepton analysis can be increased using Machine Learning. We also show how to deal with challenges in the efficiency correction and the need for additional checks for unwanted biases.

Keywords: Machine Learning; HADES; Heavy-Ion Collisions; GSI; Dileptons