Speaker
Description
Accurate jet energy reconstruction is essential for precision studies of hadronic Higgs, W, and Z boson decays at the Circular Electron Positron Collider (CEPC). The CEPC Reference Detector Technical Design Report reports a baseline jet energy resolution of approximately 4.5–5.0% for light-quark jets in the barrel region. Building on this benchmark, this study explores machine learning to exploit the information within individual jets and further improve jet energy reconstruction.
In this talk, we present an XGBoost-based jet energy calibration framework for simulated e^+ e^-→ZH events at a centre-of-mass energy of 240 GeV, using a comprehensive set of jet-level observables derived from particle-flow jets reconstructed with the Durham algorithm in FastJet. Preliminary results in a barrel benchmark sample show that the jet-count-weighted average of the double-sided Crystal Ball core resolutions across energy bins decreases from approximately 4.50% to 3.73%. The presentation will also assess the impact of the calibration on Higgs dijet invariant-mass reconstruction and its potential benefits for event selection and signal extraction in precision Higgs analyses.
