Mainz 2022 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

HK: Fachverband Physik der Hadronen und Kerne

HK 13: Heavy-Ion Collisions and QCD Phases IV

HK 13.5: Vortrag

Montag, 28. März 2022, 17:00–17:15, HK-H2

Classifying the QCD equation of state in heavy-ion collision experiments with Deep Learning — •Manjunath Omana Kuttan^1,2,3, Kai Zhou¹, Jan Steinheimer¹, Andreas Redelbach^1,4, and Horst Stöcker^1,2,5 — ¹FIAS, Frankfurt am Main, Germany — ²Institut fur Theoretische Physik, Johann Wolfgang Goethe Universität, Frankfurt am Main, Germany — ³Xidian-FIAS international Joint Research Center, Frankfurt am Main, Germany — ⁴Institut fur Informatik, Johann Wolfgang Goethe Universität, Frankfurt am Main, Germany — ⁵GSI Helmholtzzentrum fur Schwerionenforschung GmbH, Darmstadt, Germany

We present a novel technique to identify the nature of QCD transitions that happen in a heavy-ion collision experiment, particularly at the CBM experiment [1]. We show that Deep Learning (DL) models based on PointNet can be used as a fast, online method for identifying a first order phase transition from a crossover transition that happens in heavy-ion collision experiments. We use a comprehensive data preparation method to train and evaluate the models in several hypothetical experimental scenarios. A model trained on the reconstructed tracks from CBM detector simulations requires only about 40 events for accurate predictions. This makes the PointNet models an ideal candidate for online analysis of the continuous datastream produced in the CBM experiment. The DL model is shown to have up to 99.8% prediction accuracy and outperforms conventional methods based on mean observables such as the V₂ or <P_t>.

[1] Omana Kuttan, M., et al. JHEP, 2021(10), 1-25.