The ATLAS trigger is a three level system which is designed to reject background events as early as possible in the trigger chain. Events are first processed at the Level-1 trigger, which identifies interesting physics objects (e.g. jets, electrons or large MET) above a given energy threshold. At this stage only coarse information from the detector is available. In the second level trigger (Level-2) the identified objects are reconstructed using the full granularity of the detector, which improves the energy and the position resolution of the objects. Finally the third trigger level, the Event Filter (EF) has the full event available for further refinement.
Currently there is no L2 MET algorithm implemented in the ATLAS High-Level Trigger (HLT). The reason for this is the limited time budget of O(40ms) for processing an event at L2., which is a basic requirement for all Level-2 algorithms. The time constraint at Level-2 does not allow the computation of the MET with the full ATLAS calorimeters, which consists of more than 200000 channels. Two solutions are envisaged to have a L2 MET algorithm that are currently being developed by the SLAC group:The first is based on the so-called mHT. In that case the missing transverse energy is computed based on the information from jets. The jet information is obtained from the L2 algorithm. The advantage of this algorithm is that it can use the jets from the Level-2 trigger and simply adds components of the transverse momentum of the Level-2 jets. Hence it should be very fast. Furthermore it is robust, since it contains only well defined objects (Level-2 jets) problems can be pinned down more easily than when using a calorimeter cell based MET computation. The disadvantage of this method is that it is very sensitive to the Level-2 jet energy scale.
The second solution is based on the information from the LAr Front-End Boards. The Front-End Boards receive signals from 128 calorimeter channels, which are summed and create partial energy sums, Ex, Ey and Ez. This computation happens in hardware and the basic idea is to transfer only the information from the FEB instead of the individual channels. Hence, a large fraction of processing is done in hardware and more time for the actual Level-2 MET algorithm is at hand. Previous studies have shown that the performance in terms of resolution is comparable to the MET obtained based on the individual calorimeter channels. This has been tested successfully at the Event Filter level, but has not been implemented at L2 yet. The challenge for this method is the data transfer from the Level-1 trigger to the Level-2 processors.