An Hardware based Tracker for the ATLAS Experiment: Commissioning and Trigger studies

During the Run-2 of the Large Hadron Collider (LHC) the instantaneous luminosity exceeded the nominal value of $10^{34}$~cm$^{-2}$s$^{-1}$ and the number of overlapping proton-proton interactions per bunch crossing increased up to a mean value of $50$. These conditions are expected to become even worse during the future runs of the LHC, posing a challenge to the trigger systems of the experiments that have to manage rates while keeping a good efficiency for interesting physics events. In order to achieve the required online data reduction in the trigger and data acquisition system, essential to maintain the full discovery potential at such high luminosities and pileup, the LHC experiments need to optimize silicon detector information. As an example, the reconstruction of the track trajectories close to the interaction points can allow to distinguish and subtract the contributions of each pileup collision. Because of its fine resolution and granularity, tracking information is critical for distinguishing which events, selected by the first trigger level (L1), should be kept for further processing. However, extensive tracking in such environment is prohibitively expensive in terms of processing time per event or computing cores needs. Therefore, full event tracking can only be performed at low rates ($\approx$ few kHz), while the track reconstruction is performed sparingly in specific regions of interest (ROI) which have already been identified as potentially interesting by the L1 trigger. This approach has limitations in several cases. Firstly, there is a limit to either the number or size of ROIs processed by the High Level Trigger (HLT), which forces additional non-tracking cuts to be applied, resulting in reduced efficiency or higher thresholds for the objects considered. Secondly, there are cases where global event information, such as the location of the hard interaction vertex or the number of primary vertices in the event, are useful for object selections or corrections to other detector quantities. Both these problems are particularly critical for the trigger selection of signatures containing third generation fermions, such as $\tau$ or $b$-jets, for which tracking information are fundamental to keep the selection performances high. In order to facilitate the use of tracks in the HLT, the ATLAS experiment planned the installation of a hardware processor dedicated to online tracking: the Fast Tracker (FTK). It is a hardware based tracking system, currently in its commissioning phase, designed to perform full scan tracking at the full L1 trigger rate ($100$ kHz), providing reconstructed tracks to the ATLAS HLT in a mean latency of about $100$~$\mu$s, adequate for the online trigger selection. FTK is a very complex system, composed of approximately $450$ electronic boards based on two different standards (VME and ATCA). It counts about $10000$ links, including internal interconnections and connections to the other ATLAS subsystems. FTK exploits Associative Memory (AM) and Field Reprogrammable Gate Array (FPGA) for the pattern recognition and track fitting procedures, allowing for a huge parallelization of the tracking process. The custom chips and electronic boards are developed by a consortium of $10$ institutes. The final system will be composed of $8000$ memory chips and $2000$ FPGAs. This thesis work is focused on the FTK project, and especially on its commissioning phase. The first chapter of this thesis is dedicated to an introduction of the LHC and ATLAS experiment. In the second chapter, the tracking problem at the high energy physics experiments is treated, with particular focus on the solutions that the various experiments have developed (or are developing). In the third chapter, an introduction to the FTK system is provided. FTK is described in some details, focusing on the problems that we encountered during the commissioning of the system and the adopted solutions, together with an updated status of the art. In the part of the thesis describing my personal contributions, three main topics are treated. In chapter four, the results of a study on the different boards power consumptions is presented, together with a full characterization of the FTK cooling system. This study was particularly critical in order to prove the ability of the FTK housing infrastructure to cope with the very high power dissipation of the FTK boards. In chapter five, the FTK online software is presented. Particular focus is given to the peculiar requirements that a complex system as FTK pose to the development of the online software infrastructure. The problems we encountered during the development of the software framework, as well as the issues we had to face for integrating FTK into the ATLAS common infrastructure, are presented. The last chapter is dedicated to a study meant to the creation of a new single-$\tau$ trigger chain, able to exploit the FTK characteristics. In particular, a trigger chain, able to increase the signal acceptance for the search of the $H^+$ charged Higgs boson, predicted by many Beyond Standard Model theories, is presented. This trigger chain exploits the FTK tracks to increase the $\tau$ selection efficiency at low $p_T$ values, allowing to increase the signal acceptance for the $H^+\rightarrow \tau \nu$ physics channel.