guest ::
| Home > Publications database > Assessing the limits of resolution in top quark pair reconstruction algorithms |
| Home > Publications database > Assessing the limits of resolution in top quark pair reconstruction algorithms |
| Bachelor Thesis | PUBDB-2025-05377 |
; ;
2025
Abstract: This thesis studies the reconstruction of top quark-antiquark (t¯t) pairs in the dilepton decay channel, where both top quarks decay leptonically and two neutrinos escape detection. Since the event kinematics are underconstrained, the Sonnenschein algorithm, a fully analytical method used by the CMS and ATLAS experiments at the LHC, where quark-antiquark pairs of the heaviest elementary particle are produced in proton-proton collisions, is applied. The analysis focuses on the reconstructed invariant mass of the t¯t system and the factors limiting its resolution. A systematic, stepwise study is carried out using simulations of increasing realism: from parton-level events generated with Powheg, through showering and hadronization with Pythia 8, up to full detector simulation with Geant4. This approach quantifies how physics and detector effects degrade the reconstruction and highlights the intrinsic limitations of the Sonnenschein method.The results demonstrate that the usage of the Sonnenschein algorithm already introduces some mis-reconstruction even with idealized inputs, reaching a relative bias of 0.1% and a relative resolution of 2.5%. This can be attributed to numerical instabilities of the algorithm itself. Furthermore, its performance degrades significantly under realistic conditions. Transitioning from parton-level to detector-level inputs introduces substantial information loss, particularly from parton showering and jet formation, leading to resolutions exceeding 25% in high-mass regions and notable biases originating mostly from the constraints of the equations and event topologies. The standard practice of selecting the lowest-mass solution is shown to induce systematic biases, especially at higher masses.Mitigation strategies, including mass sampling from Breit–Wigner distributions and smearing techniques, improve the reconstruction efficiency and reduce resolution to an average of 23.7% at detector level, highlighting the algorithm’s sensitivity to input imperfections. These findings indicate that while the Sonnenschein algorithm provides a robust analytical foundation for top-pair reconstruction, further improvements, such as refined jet detection strategies and machine learning-assisted solution selection, are essential for exploiting its full potential in realistic experimental contexts.
|
The record appears in these collections: |
PDF
PDF (PDFA)