Design and performance of the ATLAS jet trigger system

The CERN Large Hadron Collider is the biggest and most powerful particle collider made by man. It produces up to 40 million proton-proton collisions per second at unprecedented energies to explore the fundamental laws and properties of Nature. The ATLAS experiment is one of the detectors that analyse and record these collisions. It generates a huge data volume that has to be reduced before it can be permanently stored. The event selection is made by the ATLAS trigger system, which reduces the data volume by a factor of 10^{5}. The trigger system has to be highly configurable in order to adapt to changing running conditions and maximize the physics output whilst keeping the output rate under control. A particularly interesting pattern generated during collisions consists of a collimated spray of particles, known as a hadronic jet. To retain the interesting jets and efficiently reject the overwhelming background, optimal jet energy resolution is needed. Therefore the Jet trigger software requires CPU-intensive reconstruction algorithms. In order to reduce the resources needed for the reconstruction step, a partial detector readout scheme was developed, which effectively suppresses the low activity regions of the calorimeter. In this paper we describe the overall ATLAS trigger software, and the jet trigger in particular, along with the improvements made on the system. We then focus on detailed studies of the algorithm timing and the performance impact of the full and partial calorimeter readout schemes. We conclude with an outlook of the jet trigger plans for the next LHC data-taking period.