Heidelberg 2022 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 70: Experimental Methods (general) 3
T 70.9: Vortrag
Mittwoch, 23. März 2022, 18:15–18:30, T-H29
Towards tuning electromagnetic shower properties to data with AtlFast3 — •Joshua Beirer1,2, Michael Duehrssen1, and Stan Lai2 — 1CERN — 2Georg-August-Universität Göttingen
AtlFast3 is the next generation of high precision fast simulation in ATLAS and encompasses a parametrised and a machine-learning approach based on Generative Adversarial Networks (GANs). With respect to its predecessor, AtlFast3 significantly improves in physics performance while retaining the benefit of a considerably faster simulation in comparison to Geant4.
A precise simulation of electromagnetic (EM) shower properties in the ATLAS calorimeter is crucial for the identification of particle showers originating from electrons and photons. While AtlFast3 precisely simulates the properties of EM showers, it inevitably inherits any mismodelling of the full Geant4 simulation, upon which its parametrization is based. Differences between the Geant4 simulation and data collected by the ATLAS detector are well known but insufficiently understood. Traditionally, these discrepancies are corrected using ad hoc methods such as the applications of shifts to the central values of the corresponding distributions, a procedure known as fudging.
In this talk, a brief overview of fast simulation in ATLAS is given. Furthermore, the development of different models directly embedded within the simulation framework used to tune EM shower properties directly to data are described and it is shown that AtlFast3 can be modified in a way that the shower shapes observed in data are accurately reproduced by the simulation.