Göttingen 2025 – wissenschaftliches Programm

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

T: Fachverband Teilchenphysik

T 2: Higgs Physics I (HH and trilinear coupling)

T 2.4: Vortrag

Montag, 31. März 2025, 17:30–17:45, ZHG104

Neural-network-based di-tau mass reconstruction in Higgs boson pair production in the final state with two b quarks and two tau leptons — •Jonathan Pampel, Tatjana Lenz Lenz, and Jochen Dingfelder — Physikalisches Institut, Universität Bonn, Nussallee 12, 53115 Bonn

The Higgs boson self interaction could not yet be observed at the Large Hadron Collider due to the rarity of associated processes, such as Higgs boson pair production. Upper limits on the Higgs self-coupling strength have been set using ATLAS and CMS pp data from LHC Run 2. Run 3 data will improve the limits on the HH production cross section and on the Higgs self coupling.

Tau leptons provide a relatively distinct signature (triggering) during data taking and with a probability of about 6% for Higgs bosons to decay into tau pairs, this process is rather frequent. However, the most abundant decay mode for Higgs bosons is the decay into two b quarks. The HH->bbττ decay mode benefits from both advantages.

One of the challenges of studying this decay mode is the reconstruction of the invariant mass of the di-tau system. This has long been done using a fitting tool -- the missing mass calculator (MMC) -- which performs well, but is computationally expensive and sometimes does not converge. To mitigate this issue, a neural network (NN) can be used since its evaluation is faster and there is no convergence issue.

This talk will present the training and the performance of the NN-based method for di-tau mass reconstruction, applied to ATLAS pp collision data from Runs 2 and 3.

Keywords: Higgs; MMC; Tau; NN; DiTauMass