Wing Sheung Chan

68 Event selection and classification malfunctions or glitches happen. To ensure that events recorded when the detector is at fault can be identified and discarded in analyses, the conditions of the detector are routinely monitored by human operators and automated algorithms during and after data taking [91] . They are stored as data quality (DQ) information alongside each recorded event. For the search for Z → `τ decays, selected events are required to be recorded without any critical malfunction in any of the subsystems of the detecto r † . Main sources of data loss include data corruption due to desynchronisation of the read-out electronics, high-voltage trip s ‡ due to failures of the power supplies and noise burst s § . With the continuous firmware and software development, as well as the replacement of the read-out hardware, these conditions have been improved over the course of the data taking. Overall, the complete Run 2 data set has a combined DQ efficiency of 95.6%. In addition to the data quality requirement, events are required to have a primary vertex with at least two associated tracks. The primary vertex is selected as the vertex with the largest P p 2 T , where the sum is over all tracks with transverse momentum p T > 0 . 4 GeV that are associated with the vertex. 4.1.2. Triggers In the analysis, only events that passed the unprescaled single-lepton HLTs summarised in Table 4.1 are used. For the data taken in 2016-2018, these triggers require at least one electron (or muon) in the event to have p T > 26 GeV if it is isolated, or p T > 60 GeV ( 50 GeV ) if it is not isolated. For the data taken in 2015, slightly looser thresholds are employed by the triggers. Since the electron or muon from a Z → `τ decay can always be expected to be boosted and isolated, the selected triggers have a high efficiency for the signal events. Table 4.1.: High-level triggers used in the analysis. Detailed definitions and performance of the triggers can be found in Refereces [92, 93] . 2015 2016-2018 Single-electron triggers HLT_e24_lhmedium_L1EM20VH HLT_e26_lhtight_nod0_ivarloose HLT_e60_lhmedium HLT_e60_lhmedium_nod0 Single-muon triggers HLT_mu20_iloose_L1MU15 HLT_mu26_ivarmedium HLT_mu40 HLT_mu50 † Except events that only encountered errors at the HLT or data-acquisition level, where information can be recovered offline. These events enter the so-called debug stream, and are included in physics analyses to avoid missing potentially interesting events. They constitute a fraction of less than 10 − 7 of all the recorded events in Run 2. ‡ A sudden drop of high voltage that makes the energy correction unreliable. § A burst of coherent noise over a large region of the detector that typically lasts a microsecond. It mainly affects the LAr end-caps.