ATLAS Track and Vertex Reconstruction for Run-3 and High-Luminosity LHC

The determination of charged-particle trajectories (tracking) and the identification of primary collision vertices (vertexing) are complex parts of the event reconstruction chain in collider experiments and constitute the building blocks of most high level analysis objects. During the Run 2 data-taking in ATLAS, tracking was by far the most resource intensive step, for an average number of p-p collisions per bunch crossing (pile-up) ranging from 20 up to 60. The complexity of the combinatorial problem increases dramatically with pile-up and the physics performance degrades as more low-quality tracks with mis-assigned, missing or randomly combined hits are reconstructed. Averages of around 50 interactions per bunch-crossing are expected during the LHC Run 3, rising to about 200 during the High-Luminosity (HL) phase of the LHC, scheduled to start in about 5 years. In order to cope with these challenging conditions and to maintain the physics performance reached up to LHC Run 2, a major rewrite of the Run 3 reconstruction software was performed while ATLAS prepares for a replacement of the current ATLAS Inner Detector with a new all-silicon Inner Tracker (ITk) for HL-LHC. The Run 3 software improvements allowed to dramatically increase the reconstruction speed and pileup robustness. This included replacing the existing ATLAS vertexing with the pioneering use of elements of the ACTS software framework, which will become the backbone of ITk track reconstruction, in production. In this talk, the improvements achieved for the track and vertex reconstruction to be used in the upcoming LHC Run 3 as well as the latest results on the expected performance of the ITk tracking and of other high-level object identification will be presented.