Wing Sheung Chan

Signal and background modelling 87 samples are generated for top-fusion productions, which have negligible contribution to the total background in the relevant regions. The overall yields of the Higgs boson samples are normalised using the predicted cross sections and Higgs-decay branching fractions recommended by the LHC Higgs Cross Section Working Group in Reference [118] . The cross sections are calculated at next-to-next-to-next-to-leading-order (N3LO) in α s and NLO in the electroweak coupling constants. After being generated, all simulated events are set to go through a detailed simulation of the ATLAS detector implemented using Geant 4 [119] . Furthermore, simulated inelastic pp collisions, generated using Pythia 8 with the NNPDF2.3 LO PDF set and the A3 tune [120] , are overlaid with the simulated hard-scattering events to model pile-up. The events are reweighted to match the pile-up conditions measured in the different data taking periods. Finally, all simulated events are processed using the same reconstruction algorithms as the ones used for the actual data, which have been introduced in Chapter 3. 5.2. τ polarisation reweighting As mentioned earlier, the simulated signal events are generated with unpolarised τ lep- tons (nominal signal), which corresponds to the assumption of a parity-conserving Z`τ interaction vertex. However, depending on the mechanism behind the LFV coupling, the effective Z`τ vertex could be parity-violating and the τ leptons from the Z -boson decays could be polarised. The τ polarisation state determines the decay topology of the τ lepton, which in turn affects the kinematic distributions of the reconstructed τ had - vis and E miss T in a signal event. Therefore, the assumption for the τ polarisation has a significant impact on the acceptance of τ had - vis candidates and the NN output distributions for signal events in the SR. This ultimately impacts the signal sensitivity of the search. In order to generalise the analysis, the nominal signal samples are reweighted using the software package TauSpinner [121] to consider as well the scenarios in which the effective Z`τ vertex is maximally parity-violating and the τ leptons from the LFV Z -boson decays are either all left-handed or all right-handed. Different constraints on the LFV branching fraction can then be set for the different considered scenarios in case of no discovery. For each generated signal event, TauSpinner uses the truth event records to identify the τ decay mode and the kinematics of the final-state particles. It then calculates the probabilities of occurrence of the event, averaged over all possible initial states according to the assumed parton distribution (NNPDF2.3 LO), under the assumptions that the τ lepton is left-handed or unpolarised. By dividing the probability under the left-handed τ assumption by the probability under the unpolarised τ assumption, TauSpinner obtains and outputs an event weight w ( p = − 1) that can be used to reweight the nominal signal sample to simulate the scenario where only left-handed τ leptons are produced.