Triggering on hadronic tau decays in ATLAS: semiconductor tracking detectors in action

Identifying the decay of the hadronic tau leptons plays a crucial role in the search for physics beyond the Standard Model as well as in Standard Model measurements. However, these decays are di ﬃ cult to identify and trigger on due to their resemblance to QCD jets. Given the large production cross section of QCD processes, designing and operating a trigger system with the capability to e ﬃ ciently select hadronic tau decays, while maintaining the rate within the bandwidth limits, is a di ﬃ cult challenge. This contribution will summarize the status and performance of the ATLAS tau trigger system during the 2011 data taking period, emphasizing the key role of semiconductor tracking detectors for tracking and vertexing. Di ﬀ erent methods that have been explored to obtain the trigger e ﬃ ciency curves from data will be shown. Finally, in light of the vast statistics collected in 2011, future prospects for triggering on hadronic tau decays in this exciting new period of increased instantaneous luminosity will be presented.


Introduction
The ATLAS trigger system [1] is aimed at reducing the initial  towers. Both the core and isolation energy at L1 have been 45 studied and validated in detail and found to match our simula-46 tion as seen in Fig. 1 Figure 1: Resolution of the L1 cluster energy with respect to the p T of offline tau candidates. The L1 object is required to be associated to the EF tau20 medium1 chain and match to an offline probe tau within ∆R < 0.2. The offline candidates are selected using a tag and probe analysis with Z → ττ → µh events collected by ATLAS during Summer 2011. The analysis follows closely the Tau ID efficiency measurement [3].  Figure 2: Isolation energy of L1 objects associated to the EF tau20 medium1 chain and matched to offline probe taus within ∆R < 0.2. The offline candidates are selected using a tag and probe analysis with Z → ττ → µh events collected by ATLAS during Summer 2011. The analysis follows closely the Tau ID efficiency measurement [3].

61
The high level trigger, HLT, has access to information from 62 all the subdetectors with full granularity, but only in the RoI 63 around the candidate that fired L1. A dijet selection has been applied to select the events in data.

77
The RoIs that passed the L2 requirements are reprocessed at       mentum threshold of 20 GeV is shown in Fig. 6. In both cases 136 the efficiencies are very high, 100% in EF and 99% and 98% 137 in L2 in the plateau region for muon and electron efficiency re-138 spectively.
139 Figure 6: L2 and EF Si-tracking efficiency for offline muon tracks of |η| < 2.5 and that are located inside L2 muon regions-of-interest, shown as a function of the transverse momentum of the offline muon. Offline tracks are selected by cuts that would reduce secondaries and choose tracks which have at least a minimum number of Si hits: |η| < 2.5, N pixelhits > 0, N S CT clusters > 5, |Z o f f lineVertex | < 200 mm, with respect to the offline vertex: |a 0 | < 1.5 mm, |∆z sin θ| < 1.5 mm.

140
The tau trigger efficiency has been studied with different 141 methods. Physics analysis using data collected with a tau trig-142 ger need to understand the trigger efficiency in order to correct 143 properly the estimations from the simulation samples.
been performed with a tag and probe method [3], using data  the instantaneous luminosity is shown in Fig. 8 and 9     : Event Filter output rates versus the instantaneous luminosity measured by ATLAS for selected HLT tau chains collected from September to October 2011. The numbers in the item names correspond to E T requirements at EF for a given trigger type, e.g. an electron candidate with E T > 15 GeV in e15vh, where vh implies that the η dependent E T selections were applied at L1 and the energy in core of the hadronic calorimeter is required to be smaller than 1 GeV. The data have been collected at a center-of-mass energy of 7 TeV in 2011.
In order to keep the efficiency for true hadronic tau decays 184 within the limits of the assigned bandwidth, some upgrades are 185 being implemented. These upgrades will include the use of 186