Dortmund 2021 – scientific programme

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

T: Fachverband Teilchenphysik

T 28: Top quark production II

T 28.4: Talk

Tuesday, March 16, 2021, 16:45–17:00, Tc

Systematic-aware top-quark pair reconstruction with deep learning — Tomas Dado, Johannes Erdmann, •Lars Kolk, and Olaf Nackenhorst — TU Dortmund University

The top quark plays a unique role in the Standard Model of particle physics as it is the most massive of all known elementary particles. Due to its large Yukawa coupling, a precise measurement of its properties is crucial in order to search for hints for physics beyond the Standard Model. Since the average lifetime of a top quark is smaller than the hadronisation timescale, it decays before it can form a bound state.

In the Standard Model, the top quark almost exclusively decays into a bottom quark and a W-boson. The W-boson can then either decay into a charged lepton and its respective neutrino or into an up- and a down-type quark. In this work, t t production with one charged lepton in the final state is studied, which results in four jets, two of which are b-jets, at leading order. In order to calculate the four momenta of the top quarks, the detected jets must be assigned to the final state particles of the hard scattering process, using the kinematic properties of the decay products. This process is called tt reconstruction.

Deep Neural Network (DNN) have shown to outperform commonly used algorithms in tt reconstruction with one charged lepton in the final state (J. Erdmann et al 2019 JINST14 P11015). In this work, the DNN approach is modified in order to minimise the impact of modelling uncertainties, which many top quark analyses suffer from. Initial studies of the top quark reconstruction using DNNs using Monte Carlo simulated samples are presented.