Giannini, Antonio (2020) Machine Learning methods for diboson searches in semi-leptonic final states with the ATLAS experiment at LHC. [Tesi di dottorato]


Download (22MB) | Preview
[error in script] [error in script]
Item Type: Tesi di dottorato
Resource language: English
Title: Machine Learning methods for diboson searches in semi-leptonic final states with the ATLAS experiment at LHC
Date: 12 March 2020
Number of Pages: 248
Institution: Università degli Studi di Napoli Federico II
Department: Fisica
Dottorato: Fisica
Ciclo di dottorato: 32
Coordinatore del Corso di dottorato:
Capozziello, SalvatoreUNSPECIFIED
Merola, LeonardoUNSPECIFIED
Carlino, GianpaoloUNSPECIFIED
Date: 12 March 2020
Number of Pages: 248
Settori scientifico-disciplinari del MIUR: Area 02 - Scienze fisiche > FIS/01 - Fisica sperimentale
Date Deposited: 31 Mar 2020 16:13
Last Modified: 08 Nov 2021 12:22

Collection description

Beyond Standard Model (BSM) Physics is one of the topic in the LHC search program. The collision energy of 13 TeV in the center of mass could be produce new physics phenomena that have not been observed yet. The Standard Model of particles physics is a good model that describes many of the phenomena knows today but does not give explanations for some phenomena as the Dark Matter observations and neutrino masses; furthermore, this model is not completely satisfying by a theoretical point of view since the several free parameters that have to be measured. Many BSM searches are ongoing in the ATLAS collaboration. In this scenario, several new models predict the presence of new particles in the high mass range, with different spins (2HDM, HVT, KK-Graviton) and production mechanisms. The dibosons channel decay of this new particle (X—>VV, V=Z,W) is one of the most relevant signature. A semi-leptonic final state is a good compromise in order to have a clear signature with the leptons coming from one boson and a decent Branching Ratio. My thesis focuses on the study of diboson semi-leptonic final states. This final state could be interpreted both with a search of a new particle decaying in diboson both in a non-resonant interpretation looking for an extremely rare process, predicted by the Standard Model, Vector Boson Scattering (VBS). These interpretation are both developed with the introduction of Machine Learning approaches. A Recurrent Neural Network is introduced for the categorisation of the production mechanism of the new resonance, Vector Boson Fusion (VBF) VS gluon gluon Fusion (ggF), using the jets informations; this kind of network is used for the first time in ATLAS to identify a VBF topology. A challenging development has been performed in the direction of Machine Learning techniques besides a significant improvement in the VBF selection up to 20-60% depending on the mass of the resonance. Exclusion limits on the production cross section of are resonances have been derived. A BoostDecisionTree (BDT) is used to improve the separation of the VBS process from the others Standard Model background and to use this variable a final discriminant to perform the cross section measurement of this process. A measurement of the cross section has been derived and found in agreement with the Standard Model expectation.


Downloads per month over past year

Actions (login required)

View Item View Item