Wing Sheung Chan

Object reconstruction and identification 59 3.3. Electrons In the ATLAS detector, electrons typically appear as a single track in the inner detector that matches geometrically to a distinctive shower in the EMCal [83– 85] . The reconstruction of electrons begins with the reconstruction of clusters in the EMCal. A specialised cluster building algorithm is used for electron reconstruction [85] . It creates dynamic, variable-sized clusters, known as superclusters, that are able to recover energy from bremsstrahlung photons or from electrons from photon conversions adaptively. An algorithm is then used to associate tracks to the clusters to form electron candidates. A track is associated if it has p T > 1 GeV and is within an angular distance of ∆ R = 0 . 3 to the centre of a cluster. If multiple tracks are found, only the track that leaves hits in the SCT and is closest to the centre of the cluster are chosen. Since electron and photon showers are basically identical, the presence of an associated track is essential in telling an electron and a photon apart. Therefore, the reconstruction of electrons is restricted to the central region | η | < 2 . 47 , which is within the coverage of the inner detector. Finally, the energy of the electron candidate, calculated from the energy deposits in the EMCal, is calibrated using studies with Monte Carlo simulations and observed Z → ee events. Similar to τ had - vis , the reconstruction of electrons is susceptible to misreconstructed jets. A multivariate likelihood-based identification is used to improve the purity of reconstructed electron candidates. The identification uses information from both the inner detector and the calorimeters. The information used includes leakage to the HCal, lateral shower width and longitudinal energy distribution in the EMCal, track conditions, TRT response and track-cluster matching. The probability density functions of the relevant quantities are multiplied together to form the likelihood functions for electron signals ( L S ) and for jet backgrounds ( L B ). Then, a discriminant d L = L S L S + L B (3.8) is calculated. The probability distribution of d L has a property of peaking sharply at zero for signals and at one for backgrounds. The value of the discriminant is then used to select electron candidates. Three WPs, Tight , Medium and Loose , are defined by cuts on the discriminant that correspond to increasing identification efficiencies. Another challenge for selecting electron candidates is to draw a distinction between electrons produced directly from hard-scattering vertices or decays of heavy resonances (prompt electron) and those produced from decays of quarks or hadrons within a jet. The main difference between the two types of candidates is in the activity around the reconstructed electron. Two types of isolation variables are calculated. Calorimeter-based isolation variables are calculated based on the total energy of all clusters within a certain distance from the electron candidate, but with the contribution of the candidate itself being subtracted. Track-based isolation variables are constructed based on the total transverse momentum of all tracks around the track associated to the electron candidate, excluding the associated track itself. Various isolation WPs are defined to cater for the different