Olga Sunneborn Gudnadottir: Of dark mesons and novel methods: A dark sector search in ATLAS data and development of new techniques for challenging final states
- Date: 30 May 2024, 13:15
- Location: Sonja Lyttkens (101121), Lägerhyddsvägen 1, hus 10, Uppsala
- Type: Thesis defence
- Thesis author: Olga Sunneborn Gudnadottir
- External reviewer: Markus Klute
- Supervisors: Rebeca Gonzalez Suarez, Raazesh Sainudiin, Richard Brenner, Jochen Jens Heinrich
- Research subject: Physics
- DiVA
Abstract
Studies of the interactions of elementary particles at high energies have been carried out at the Large Hadron Collider (LHC) at CERN for over a decade. Different quantities from the Standard Model (SM) of particle physics have been measured with increasing accuracy without substantial deviations from predictions. Searches for physics beyond the SM are similarly carried out, motivated by the existence of phenomena not yet described by it, such as dark matter. This thesis presents one such search in proton-proton collision data recorded by the ATLAS detector. The search is guided by a new, proposed addition to the SM, where the dark matter candidate arises as a composite particle of a new sector. If this were realised in nature, the same sector would give rise to other composite particles, dark mesons, that would be produced in proton-proton collisions and decay promptly to SM particles. This new model is largely free of previous constraints from searches and measurements. The full analysis targeting pair produced dark pions decaying to top and bottom quarks, tttb or ttbb, in the 1-lepton channel is described. It is carried out in the full Run 2 dataset of 140 fb−1 of proton-proton collisions at √s = 13 TeV center-of-mass energy. The analysis is sensitive to large parts of the parameter space of the model, and no significant excess was seen over SM predictions. Based on this, limits on the production cross-section of dark pions were set. By comparing with the theoretical cross sections of the model, these rule out dark pion masses up to 943 GeV in the most sensitive configuration.
Further, several novel techniques that could aid with searches in similar phase-spaces are presented. First, the Extrapolation Engine fast simulation of the inner tracker for the high luminosity upgrade of ATLAS was used in the study of a proposed hardware track trigger (HTT). This could be crucial to retaining efficiency in similar phase-spaces in the extreme conditions at the high luminosity LHC (HL-LHC). Second, the fully scalable multi-dimensional density estimate in SparkDensityTrees was applied on background and signal similar to those in the dark meson analysis and was shown to efficiently find signal-enriched regions. Third, the unsupervised clustering algorithm UCluster which can be trained with any clustering objective, such as signal extraction, anomaly detection or jet tagging was developed to run on multiple cores for arbitrary scalability. Lastly, a Boosted Decision Tree (BDT) was applied for signal and background discrimination in the dark meson analysis, yielding promising results for future iterations of it.