Electronic Theses and Dissertations (PhDs)

Permanent URI for this collectionhttps://hdl.handle.net/10539/38021

Browse

Now showing 1 - 2 of 2

Implementation of the DAQ software for the ALTI module in the ATLAS TileCal and the search for new physics in the four lepton final state
(University of the Witwatersrand, Johannesburg, 2023-06) Tlou, Humphry Sijiye; Wilkens, Henric; Ruan, Xifeng; Mellado, Bruce
The discovery of the Standard Model (SM) Higgs boson in 2012 presents new challenges and opportunities for the Large Hadron Collider (LHC) experiments. After a long period of operation, the LHC experiments needed to maintain and upgrade their detectors in order to continue and conduct research beyond the SM. As part of the upgrades, the Tile Calorimeter (TileCal) participated in Phase-I of the upgrades (December 2018 - March 2022). TileCal, the central hadronic calorimeter (|η| < 1.7) of the ATLAS experiment uses a set of Trigger and Data Acquisition (TDAQ) software to readout, transport and store physics data resulting from collisions at the LHC. In preparation for the Phase-I upgrades, the ATLAS Local Trigger Interface (ALTI) module was designed for the ATLAS experiment at CERN for TDAQ purposes. It is a 6U VME electronics board, which is a part of the Timing, Trigger and Control (TTC) system. It integrates the functionalities of four legacy modules, currently used in the experiment: Local Trigger Processor, Local Trigger Processor interface, TTC VME bus interface and the TTC emitter. The ALTI module provides the interface between the Level-1 Central Trigger Processor and the TTC optical broadcasting network to the front-end electronics of each of the ATLAS sub-detectors. This thesis discusses the development, validation and integration of the TileCal specific ALTI software in the TileCal online software by the author. A set of ALTI boards were installed in the back-end electronics of the sub-detector and fully validated for the ATLAS detector at CERN. Performance testing and maintenance of the ALTI modules and software were performed during the second half of the upgrade period, in preparation for Run 3 (2022–2025) data-taking period. The thesis also discusses the search for the presence of a new heavy resonance produced via gluon-gluon fusion and decaying to the four-lepton (4ℓ) final state, in association with missing transverse energy (EmissT), with ℓ = e, µ (where ℓ is the lepton, e is the electron and µ is the muon). The search uses 2015–2018 proton-proton collision data at √s = 13 TeV, corresponding to an integrated luminosity of 139 fb−1, collected by the ATLAS detector. The data are interpreted in terms of two models, firstly the R → SH → 4ℓ + EmissT , where R is a scalar boson, which decays to two lighter scalar bosons (S and H). The S decays to a pair of neutrinos and the H decays into 4ℓ, through ZZ bosons. The second model is the A → Z(νν)H(ZZ) → 4ℓ + X, where A is considered to be a CP-odd scalar which decays to a CP-even scalar H and the Z boson. The Z boson decays to X, which can be a pair of neutrinos or jets, and the H decays to the 4ℓ final state.
The application of weakly supervised learning in the search for heavy resonances at the LHC
(University of the Witwatersrand, Johannesburg, 2023-06) Choma, Nalamotse Joshua; Ruan, Xifeng; Mellado, Bruce
The discovery of the Higgs boson at the Large Hadron Collider by the ATLAS and CMS experiments has made the search for new physics beyond the Standard Model a priority in the field of High Energy Particle Physics. New resonances have yet to be discovered using inclusive and model-dependent searches, which means they may be driven by subtle topologies. Rapid improvements in Machine Learning techniques have led to their increasing application in High Energy Particle physics. Unlike supervised learning, which is known to assume full knowledge of the underlying model, semi-supervised learning, in particular weakly supervised learning, allows the extraction of new information from data with partial knowledge. The goal of this study is to set up searches for heavy resonances at the electroweak scale with topological requirements performed in both inclusive and exclusive regions of phase-space tailored to a particular production mode. These resonances could be generated with different production mechanisms. In this work, we describe search procedures based on weakly supervised learning applied to mixed samples and used to search for resonances with little or no prior knowledge of the production mechanism. This approach has the advantage that sidebands or control regions can be used to effectively model backgrounds without relying on models. The effectiveness of this method is measured by the production of the Standard Model Higgs boson, which decays into a pair of photons in both inclusive and exclusive regions of phase-space at the LHC. Having confirmed the ability of the method to extract various Standard Model Higgs boson signal processes, the search for new phenomena in high mass final states will be set up at the LHC. Subsequently, the approach is used in the search for new resonances in the Zγ final state with Z → e +e − or Z → µ +µ −, using the Monte Carlo simulated signal samples for 139 fb−1 of integrated luminosity for Run 2 collected at the LHC. The weakly supervised learning approach is implemented and compared to the performance of the fully supervised approach, which is then used to calculate the production limit for Higgs-like particles for Zγ where the significance of the signal is maximal.