Track Reconstruction at the First Level Trigger of the Belle II Experiment

Sumitted to PubDB: 2018-04-19

Category: Phd Thesis, Visibility: Public

Tags: -

Authors Christian Kiesling, Sara Pohl
Date Jan. 1, 2018
Belle II Number BELLE2-PTHESIS-2018-001
Abstract The Belle II experiment is an upgrade of the Belle experiment, which was instrumental in confirming the Kobayashi-Maskawa theory for the origin of CP violation. The upgraded e+ e− collider SuperKEKB is designed to achieve an instantaneous luminosity of 8×10^35 cm−2 s−1, which is 40 times higher than the world record set by its predecessor KEKB. Belle II will perform precision measurements in the B meson system as well as searches for rare decays, such as lepton flavor violating τ decays. As a consequence of the increased luminosity, the first level trigger has to cope with an increased background rate. A dominant background source are particles that do not originate from e+ e− collisions, but from intra-beam interactions or scattering on residual gas in the beampipe. The corresponding tracks are characterized by their production vertex, which is longitudinally displaced with respect to the interaction point. The focus of this thesis is the 3D reconstruction of tracks in the central drift chamber of the Belle II detector at the first trigger level. The drift chamber contains wires of two different orientations: axial wires are oriented along the z-axis, while stereo wires are skewed with respect to the z-axis. Assuming that tracks come from the z-axis, but not necessarily the interaction point, a track is parametrized by four parameters: the azimuth and polar angles at the track vertex, the track curvature and the longitudinal coordinate of the production vertex (“z-vertex”). The track curvature is caused by a constant magnetic field of 1.5 T that is oriented along the z-axis, and is inverse proportional to the transverse momentum of the particle. The track reconstruction in the trigger consists of two steps: a track finding step in the transverse x-y-plane and a 3D reconstruction step on individual tracks. The purpose of the track finding is to determine the number of tracks and the azimuth angle and curvature of each, using a Hough transformation. Hits on axial wires correspond to points in the transverse plane and are transformed into curves in a Hough plane, where tracks are found as the crossing points of several curves. The track parametrization for the transformation is chosen such that it represents half circles in the transverse plane, going outward from the origin. With this definition, the crossing point coordinates in the Hough plane give the azimuth angle and the signed curvature, which is related to the transverse momentum and the charge of the track. The purpose of the 3D reconstruction is to determine the polar angle and the z-vertex of each track. As the tracks have already been separated in the track finding step, each track is reconstructed independently. By combining the hits from stereo wires with the transverse track, longitudinal hit coordinates can be obtained. To increase the spatial resolution, the drift times of hits are included, which are proportional to the distance between the track and the wire. Instead of an analytical reconstruction, neural networks of the Multi Layer Perceptron (MLP) type are trained. MLPs are capable of learning nonlinear functions from data samples as well as of parallel execution with a deterministic runtime. The networks receive the wire coordinates relative to the transverse track and the drift times of hits as input and estimate the polar angle and the z-vertex. An average z-vertex resolution of (2.910 ± 0.008) cm is achieved for single muon tracks, which allows to suppress most of the displaced background tracks. After the optimization of both the Hough transformation and the neural networks, this thesis studies the trigger efficiency for selected B and τ events. Different track trigger conditions are compared, including a z-vertex veto and a veto on Bhabha scattering events, which form the second dominant background in the Belle II experiment. By combining these vetos, events with only one or two tracks can be triggered. For the lepton flavor violating decay channel τ → μγ, a trigger efficiency of 75% to 77% can be obtained with a pure track trigger, which is six times higher than without a 3D track reconstruction.
Conference Munich

Files