Search for Dijet Resonances with the Level-1 Topological Processor at ATLAS
Date issued
Authors
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
License
Abstract
Due to the limited bandwidth capacity of data acquisition systems, any analysis performed at a collider experiment has to rely on an efficient trigger system, which selects only the potentially interesting events. To cope with increasing data rates caused by higher instantaneous luminosities of up to 2.1⋅10³⁴ cm⁻² s⁻¹ in Run-2 of the LHC, trigger thresholds would have had to be increased, leading to both losses in signal efficiency and statistical limitations for a wide range of analyses. To avoid these losses, the FPGA-based Level-1 Topological Processor (L1Topo) has been introduced to ATLAS in Run-2. It extends the capabilities of the hardware-based first-level trigger by the ability
to perform trigger decisions based on topological algorithms, for example angular selections or mass cuts. Using L1Topo, it was possible to maintain the trigger thresholds for Run-2 and to avoid losses in signal efficiencies while keeping the data rates low, which is of paramount importance for analyses targeting low-p T signatures. This thesis reports on the firmware development and trigger performance of the topological processor during Run-2.
In addition, this thesis presents a novel approach to search for new physics in the dijet invariant mass spectrum. These resonance searches are performed by searching for a localized excess of events above a smoothly falling background of dijet events. Due to the limited available bandwidth of the data acquisition system, searches for dijet resonances suffer from statistical limitations in the mass range below 500 GeV. Using L1Topo, these statistical limitations can be fully avoided. By performing the analysis and creating invariant mass histograms directly on the first trigger level, the bandwidth limitations can be circumvented and every single collision event can be analyzed. This makes it possible to perform an analysis of the whole invariant mass spectrum using an unprecedented level of statistical precision. This thesis presents the development of the new approach and its implementation into the trigger and data acquisition system of ATLAS. Using this approach, data has been collected in 2018 with √s = 13 TeV in proton-proton collisions
corresponding to a recorded offline luminosity of 60.6 fb⁻¹ . A statistical analysis is performed to search for deviations from a data-driven background estimate. No evidence for new resonances is observed, the data is therefore used to set exclusion limits at a credibility level of 95 % on the product of the production cross section, the acceptance and the branching ratio, both for model-agnostic Gaussian resonance shapes and an axial-vector Z′ dark matter mediator.