Session X13: Computational Techniques For Event Reconstruction I

Live

The Deep Underground Neutrino Experiment (DUNE) is a next-generation long-baseline experiment that aims to answer some of the key questions about neutrino physics. Liquid argon time projection chamber (LArTPC), the central part of DUNE far detectors, is an advancing neutrino detector technology featuring low energy threshold, high spacial resolution and detailed event topology reconstruction. In a single-phase LArTPC, the recorded data are several 2-dimensional projection images of charged-particle trajectories, which consists of waveforms (time) on different channels (wire) produced by an induced signal of ionization electrons. For such a detector, a critical step is the TPC signal processing that reconstructs the original ionization electrons from the recorded 2D images. For the first time, we introduced a deep neural network (DNN) in LArTPC signal processing. This method showed significant improvements over traditional ones with data from ProtoDUNE-SP, a 1:1 scale prototype of DUNE far detector module components. In this presentation, we will report details of the deep neural network LArTPC signal processing method and preliminary results.   

Live

The single-phase liquid argon time projection chamber (LArTPC) provides a large amount of detailed information in the form of fine-grained measurements of ionization electrons and scintillation light from particle traces. The MicroBooNE detector has an active mass of 85 tons of liquid argon and is located along the Booster Neutrino Beam (BNB) at Fermilab. It will examine a rich assortment of physics topics, such as searches for a light sterile neutrino and measurements of neutrino-Argon interaction cross sections. The Wire-Cell event reconstruction is a novel tomographic event reconstruction paradigm for LArTPCs. It reconstructs topology-agnostic 3D space points based on multiple 2D projection views of the TPC activity by utilizing geometry, time, charge, and sparseness in spatial distribution to reduce ambiguity from individual 2D views. In this talk, we present the principle of the Wire-Cell 3D event reconstruction, including 3D charge imaging and charge-light matching, and its application to MicroBooNE data. We also report on the progress of the subsequent 3D tracking based on Wire-Cell 3D images.   

Live

During Run 1 of the LHC, the optimal reconstruction algorithm for large-R jets in ATLAS, characterized in terms of the ability to discriminate signal from background and robust reconstruction in the presence of pileup, was found to be anti-kt jets with a radius parameter of 1.0, formed from locally calibrated topological calorimeter cell clusters and groomed with the trimming algorithm to remove contributions from pileup and underlying event. Since that time, much theoretical, phenomenological, and experimental work has been performed to improve both the reconstruction of the jet inputs as well as the grooming techniques applied to reconstructed jets. In this work, an inclusive survey of constituent-level pileup mitigation algorithms, jet inputs, and grooming algorithms is done to study their pileup stability and ability to identify hadronically decaying W bosons within the ATLAS experiment. It is found that compared to the conventional reconstruction algorithm of large-R trimmed jets formed from calorimeter cell clusters, these methods can significantly improve both the pileup stability and background rejection in $pp$ collisions at $\sqrt{s}=$ 13 TeV with the ATLAS detector.   

Live

A fast hardware based track trigger is being developed in ATLAS for the High Luminosity upgrade of the Large Hadron Collider (HL-LHC). The goal is to provide the high-level trigger with full-scan tracking at 100 kHz and regional tracking at 1 MHz, in the high pile-up conditions of the HL-LHC (in $pp$ collisions at $\sqrt{s}=$ 14 TeV with the ATLAS detector). A method for fast pattern recognition using the Hough transform is investigated. In this method, detector hits are mapped onto a 2D parameter space with one parameter related to the transverse momentum and one to the initial track direction. The performance of the Hough transform is studied at different pile-up values (140 and 200) and compared, using full event simulation, with a method based on matching detector hits to pattern banks of simulated tracks stored in a custom made Associative Memory ASICs. The use of track stub finding and extrapolation is investigated to take advantage of the new ATLAS Inner Tracker in order to reduce the number of hit clusters considered by the system. A preliminary discussion of the resulting hit reduction and associated speedup, and any associated performance loss, will be presented.   

Live

The NOvA experiment is a long-baseline neutrino oscillation experiment primarily designed to measure the parameters that govern neutrino oscillations through the use of two functionally identical liquid scintilator detectors. The near detector, located 800 meters from the NuMI beam source at Fermilab, is exposed to a large flux of neutrinos, producing enough neutrino interactions for high precision cross section measurements. These measurements aim to minimize overall uncertainties on oscillation parameter estimates. To reduce dependence on interaction models, the NOvA collaboration has developed a Convolutional Neural Network (CNN) approach to identifying individual particles interacting in the detectors. These individually identified particles aid in efficient signal and background classification across a wide range of cross section analyses. A summary of recent developments and performance results from NOvA's CNN approach to single particle classification will be presented.   

Live

MicroBooNE is a liquid argon time projection chamber, designed to study neutrino interactions at energies of the order of 1 GeV. The identification of different particles produced in the interaction is a key requirement for MicroBooNE's physics goals. Particle Identification can be performed by looking at the pattern of energy depositions along the track, which requires a detailed understanding of the calorimetry of the detector. The goal of the calorimetry likelihood approach is to model energy deposition taking into account the resolution and the anisotropies of the detector. This technique also relies on data-driven corrections to the simulation, in order to reduce uncertainties in the different MicroBooNE measurements and will be of interest for the current and future liquid argon time projection chambers. We will present the current status of the method and its application to MicroBooNE data and simulation.   

Live

The MicroBooNE detector is a Liquid Argon Time Projection Chamber (LArTPC) located on the Booster Neutrino Beam (BNB) at Fermi National Accelerator Laboratory. One of the primary goals of the experiment is to study the low-energy excess of electron neutrino like events seen by MiniBooNE. This talk will discuss the deep-learning-based search for low-energy electron neutrino interactions within MicroBooNE. I will focus on the track reconstruction algorithm with the addition of charge and energy calibration and its effect on particle identification.   

On Demand

Wire Cell is an algorithm that can be used to reconstruct neutrino interactions in liquid argon time projection chamber (LArTPC) detectors in which the charge readout is performed by wire planes. We have performed simulations using this reconstruction algorithm for the general case of arbitrary number of wire planes with user defined geometry. This algorithm has been adapted to handle wires wrapped around the Anode Plane Assembly (APA) and can accept a two sided geometry.We use our simulation to calculate efficiency and purity of the reconstruction for some sample charge distributions that might result from neutrino interactions.