Boosting the discovery potential for the LHC Run 3 – Improved Track Reconstruction for prompt and long-lived particles in ATLAS

Searches for long-lived particles are among the most promising search channels for physics beyond the Standard Model (BSM) at the LHC. One strategy relies on reconstructing the decay products of potential neutral BSM particles decaying within the tracking detector volume. To maximise the sensitivity of such efforts, charged particle tracks need to be reconstructed even if their trajectories are not pointing back to the primary interaction point. This has so far been a limiting factor to these searches, as the default ATLAS event reconstruction applies strict pointing requirements in track reconstruction to stay within the bounds imposed by available computing resources. Non-pointing long-lived particle tracks were therefore reconstructed only in a small subset of the data, pre-selected by a range of filters not using tracking information and therefore severely reducing the signal acceptance. In preparation for LHC Run-3, a major effort was performed to improve the computational performance of ATLAS track reconstruction and reduce the number of incorrectly reconstructed track candidates (‘fakes’), also to accommodate a long-lived particle tracking step in the default reconstruction, removing the need for filtering. This talk will describe the improvements made to ATLAS track reconstruction, and demonstrate how this goal was achieved while still retaining an overall reduction in processing time.