Speaker
Description
New long-lived particles are a feature of many extensions to the Standard Model, and their unique detector signatures may elude searches for promptly decaying particles. An analysis of data collected in pp collisions at √s = 13 TeV with the ATLAS detector at the Large Hadron Collider is described, focusing on identifying jets produced by neutral long-lived particles decaying to Standard Model fermions within the ATLAS calorimeter. This analysis considers benchmark hidden sector models of neutral long-lived scalars with masses between 5 GeV and 475 GeV pair-produced by decays of mediators with masses between 60 GeV and 1000 GeV. A deep neural network is used to predict whether candidate jets were produced by a long-lived particle decay, QCD multijets, or beam-induced background, and an adversarial training is applied to minimize the impact of Monte Carlo mismodeling. A boosted decision tree is then used to discriminate between signal and background events based on the per-jet neural network outputs and event-level variables. The results of this analysis are presented using the full Run 2 (2015-2018) data set, corresponding to an integrated luminosity of 139/fb. No significant excess is observed, and upper limits are set for these signal models.
