Erlangen 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
T: Fachverband Teilchenphysik
T 67: Top Physics III
T 67.2: Talk
Thursday, March 19, 2026, 16:30–16:45, KH 00.011
Comparing Parton-Shower Models in Top-Quark Production and Reweighting Between Models with Machine Learning — •Arshil Shaikh1,2, Roman Kogler1,2, and Dominic Stafford2 — 1Universität Hamburg, Luruper Chaussee 149, 22761 Hamburg, Germany — 2Deutsches Elektronen-Synchrotron (DESY), Notkestr.85, 22607 Hamburg, Germany
Accurate modeling of parton-shower dynamics is essential for precision studies of top-quark production. In our setup, next-to-leading-order matrix-element events are generated with POWHEG, which can be interfaced with different parton-shower approaches. These include the transverse-momentum-ordered dipole shower of Pythia8 and the antenna-based shower implemented in Vincia. Their distinct approximations can lead to notable differences in jet and jet-substructure observables.
In this work, we study pp → tt → semileptonic events produced with POWHEG+Pythia8 and POWHEG+Vincia, comparing the tt system kinematics, jet characteristics, and jet-substructure variables such as N-subjettiness, generalised angularities, and energy-correlation functions. Furthermore, we use the DCTR (Deep neural networks using Classification for Tuning and Reweighting) technique to reweight Pythia8 events such that they reproduce Vincia-like distributions, without requiring the computationally expensive step of re-running the full detector simulation. This study highlights key differences between parton-shower models and demonstrates the potential of machine-learning-based reweighting to efficiently bridge them.
Keywords: Top Quark; Monte Carlo; Parton Shower; Machine Learning; Deep Neural Network