Optimisation of the Hadronic Tau Identification Based on the Classification of Tau Decay Modes with the ATLAS Detector

Hadronically decaying tau leptons play an essential role in the LHC physics program. Final states involving tau leptons are important to verify processes of the Standard Model of particle physics at the TeV scale, but are also of high interest for Higgs physics and beyond Standard Model studies, like Higgs CP measurements and $A\to Zh$ searches. Due to the high production cross section of QCD jets which are the dominant background, efficient reconstruction and identification techniques are crucial to guarantee an excellent selection of interesting physics events. Therefore, sophisticated multivariate algorithms are used. This thesis presents an advanced concept exploiting the information of reconstructed neutral and charged pions in the ATLAS detector, to access the tau decay substructure, and thus enhance the applicability of the tau identification to a broader field of physics analyses. First, several updates of the general algorithms used within the tau identification are implemented in order to provide a more reliable performance. This thesis focuses on the investigation of a pure substructure based tau identification. Starting with the recalculation of the default identification variables exploiting the dedicated substructure algorithms CellBased and EflowRec, the respective performance is evaluated. Moreover, new variables are defined based on the kinematics of the tau decays. Their impact on the rejection rate of fake taus as well as the correlations between them are studied. Hence, it is possible to recover and for certain configurations even exceed the performance of the currently implemented standard strategy. It can be proven that a pure substructure approach for the identification of tau leptons is achievable, and hence might be featured by various physics analyses.