The Large Hadron Collider (LHC) is designed to collide bunches of protons at a center-of-mass energy of 14 TeV with an instantaneous luminosity of 1034 cm-2 s-1. This means that two bunches collide every 25 ns in the center of the ATLAS detector. A single crossing leads to 25 proton-proton interactions on average but only a tiny fraction of the collisions result in a final state that is of interest to Standard Model analyses like, for example, a top-quark or W-boson mass measurement and an even tinier fraction is expected to result in final states that contain a Higgs boson or potential exotic particles. Needless to say that not all the detector information about each single collision event can be read out and stored to search it for interesting signatures later. A quick and efficient pre-selection of potentially interesting events is needed!

To this end, the ATLAS detector possesses a complex three-stage trigger system that has the daunting task of selecting about 200 interesting collision events out of the initial flood of 40 million events per second for permanent storage – a decision that has to be made within just a few seconds. At the same time the trigger has to be highly efficient to avoid missing interesting signatures that may correspond to new phenomena. An event that is not selected by the trigger is irretrievably lost for later analysis!

The ATLAS trigger system consists of three-stages that are designed to reduce the rate of incoming events step-by-step by applying tighter and more sophisticated selection criteria at each stage. The hardware-based Level-1 (L1) trigger reduces the initial rate from 40 MHz to 75 kHz taking about 2.5 μs per decision. It uses coarse-granularity information from the muon system and calorimeter to define Regions-of-Interest (RoIs) corresponding to large energy deposits in the detector. The software-based High-Level Trigger (HLT) consists of the two stages, the Level-2 trigger (L2) and the Event Filter (EF). The L2 trigger runs fast reconstruction algorithms on these RoIs using full-granularity information from all detector systems to make a refined selection within about 40 ms. It passes about 3,000 events per second on to the next trigger stage, the EF, which takes about 4 seconds to run reconstruction algorithms that are already very similar to the sophisticated algorithms used in the offline analysis and reduces the event rate to about 200 events per second that are selected for permanent storage and later analysis.

The Oxford ATLAS group is leading the development of triggers based on missing transverse momentum and has developed tau triggers for Run 2. This work continues the effort from Run 1, during which electron and boosted top triggers were also developed.

Former students

Name Topic
Katharina Behr Tracking for electron triggers, boosted top triggers
Mark Pickering FastTracKer (FTK) for tau triggers