Development and performance of the track finder for the Belle II Vertex Detector

Sumitted to PubDB: 2018-04-04

Category: Master Thesis, Visibility: Public

Tags: -

Authors Valerio Bertacchi, Francesco Forti, Eugenio Paoloni
Date Jan. 1, 2017
Belle II Number BELLE2-MTHESIS-2018-001
Abstract Belle II is the general purpose detector that will be operated at the second generation B-Factory SuperKEKB and it is currently in an advanced construction stage in KEK (Japan). SuperKEKB is a high luminosity asymmetric e^+e^- collider (peak luminosity up to 8 X 10^35 cm^-2s^-1), with the center-of-mass energy tuned on the peak of the Y(4S) resonance. It realizes an optimal environment to perform precision measurements in the heavy flavour sector and it will allow to collect 50 ab^-1 of Y(4S) events. The Belle II detector is designed to study a broad spectrum of heavy flavour physics, including the CP violation of B and D systems and the rare decay modes of B, tau and D. The improved precision that will be achieved, thanks to the increased statistic and to the high performance of the detector, may allow to observe physics Beyond the Standard Model at energy scales higher than those directly accessible. Several channels in the Belle II physics program already show tensions with Standard Model predictions that might be signatures of New Physics. In order to perform the B physics measurements with the required precision and to fully exploit the B-Factory environment it is essential to reconstruct the complete Y(4S) events including the momenta and the vertices of the charged particle tracks down to tens MeV/c and the energy neutral particles. To fullfill this demanding task, Belle II adopts a high performance silicon Vertex Detector (VXD), realized with two Pixel Detector layers (PXD) and four Silicon Strip Detector layers (SVD), and a Drift Chamber (CDC). The Belle II - Pisa group is involved in the production of the forward and backward regions SVD modules. This thesis work has been partially devoted to perform electrical qualification test on the produced forward and backward modules, in the INFN Laboratories, while the main focus of the activity has been on the reconstruction software development, as described below. The Belle II tracking software is designed to perform the pattern recognition, vertexing and fitting procedures in the online (i.e. in the high level trigger of the experiment) and offline event reconstruction. One of the prominent feature of the Belle II tracking software is its capability of efficiently reconstructing tracks down to pt=50 MeV/c using a VXD standalone track finder (VXDTF), necessary for instance to reconstruct the low momentum pions from D* decays. In this momentum range the pattern recognition is particularly hard: only few hits seen in VXD are available to reconstruct the tracks and the material effects, like energy losses and multiple scattering, significantly divert the trajectory from an ideal helix. To overcome this issue the VXDTF adopts the Sector Map concept. In this concept a lookup table called Sector Map, that contains the geometrical informations of the patterns, is used to compare the measured track information with the expected ones. The pattern recognition begins by building segments (pairs of his), then triplets (pairs of segments sharing the inner hit) and then four-hit track (pairs of triplets sharing the inner segment). To reject the random combinations of hits, at each step geometrical and dynamical filters with the information provided by the Sector Map are applied. The Sector Map is realized collecting the information from a sample of simulated events, called Training Sample, which defines a priori the Sector Map that will be used in the actual data taking. After the reduction of the combinatorial burden performed by the Sector Map filters the tracks are identified and collected by a cellular automaton. Then, a best candidate selection is performed by a Hopfield Neural Network to remove the overlap between tracks which share hits. During the period of this thesis the Belle II tracking group has been involved in a redesign of VXDTF code, in order to to improve its flexibility and maintainability. This redesign allows to develop new strategies to improve the performance of the track finder, in term of efficiency, fake rate and time consumption. One of the critical point of the VXDTF is the definition of the Training Sample, that directly affects the Sector Map complexity and then the track finding efficiency and purity together with CPU time and memory footprint. In fact, if inside the Training Sample there are tracks which are very far from helical trajectories, and therefore not trackable, the Sector Map will allow the collection of patterns that will be discarded in fitting phase. Therefore these tracks increases the time consumption and the fake rate of the track finder, without a relevant efficiency gain. One of the central point of this thesis is the development of the selection procedure of the Training Sample, with the definition of proper requirements on the simulated tracks in order identify and discard a priori the non trackable patterns, without discarding events of physical interest. One of the possible source of this class of tracks are low momentum multiple scattering effects which make the track depart from the ideal trajectory. This effect increases at low the momentum, coherently with the observed degradation of the performance of the track finder. A chance to identify these tracks come from the analysis of the track parameters in their evolution along the trajectory: these parameters should be constant along the track, and a strong variation in a certain position inside the detector is a signature of a unusually large interaction. The training sample selection shows several advantages: the fake rate, i.e. the number of random combination of hits selected by the VXDTF, is halved, the CPU time required by the VXDTF is reduced by a factor three and the fitting efficiency increases. The side effect is a small degradation of the low momentum tracking efficiency of some percent. Part of the thesis works covers the application of the redesigned track finder on the real data, with the participation to the DESY Testbeam in February 2017, where a six-layer sector of VXD has been tested on an electron beam with an energy up to 6 GeV. The purpose of the testbeam was to test the SVD and PXD sensors with the complete readout and trigger chain, and to test the tracking software after the substantial upgrades of the redesign. During the testbeam track finder performances studies has been performed together with the characterization of the SVD sensors in term of hit efficiency, with a very promising average value of 99.6%. In conclusion in my thesis work I have covered some of the most relevant aspect of the silicon tracking detector operations, from the single module electrical tests, to the optimization of tracking software, to the test of the tracker performance on beam.
Conference Pisa

Files