Track Reconstruction in the ATLAS Experiment: The Deterministic Annealing Filter

The reconstruction of the trajectories of charged particles is essential for experiments at the LHC. The experiments contain precise tracking systems structured in layers around the collision point which measure the positions where particle trajectories intersect those layers. The physics analysis on the other hand mainly needs the momentum and direction of the particle at the estimated creation or reaction point. It is therefore needed to determine these parameters from the initial measurements. At the LHC one has to deal with high backgrounds while even small deficits or artifacts can reduce the signal or may produce additional background after event selection. The track reconstruction does not only contain the estimation of the track parameters, but also a pattern recognition deciding which measurements belong to a track and how many particle tracks can be found. Track reconstruction at the ATLAS experiment suffers from the high event rate at the LHC resulting in a high occupancy of the tracking devices. Additionally one part of the inner tracker, the so-called Transition Radiation Tracker (TRT), shows a left/right ambiguity for each measurement which has to be solved during pattern recognition and track fit. A special algorithm was developed to handle noise and ambiguities during the track fit. The Deterministic Annealing Filter (DAF) extends the widely-used Kalman Filter with the concept of competition between measurements. Furthermore an annealing process is introduced to avoid local optima of measurement assignment and track fit. The DAF can therefore help to reduce limitations of the track reconstruction performance, e.g. at high noise levels and in dense jets. This algorithm was implemented in the ATLAS offline reconstruction software within this thesis. The performance of the algorithm was studied and its parameters were adjusted. Minor changes in the calculation of measurement assignments to the original proposal of the algorithm could improve the track reconstruction performance in the ATLAS TRT.