Neutrino event reconstruction in DUNE using Pandora - Leigh Whitehead ICHEP 2020 for the DUNE Collaboration
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Neutrino event reconstruction in DUNE using Pandora Leigh Whitehead for the DUNE Collaboration ICHEP 2020 31/07/20
Overview • Introduction to the Deep Underground Neutrino Experiment (DUNE) • Introduction to Pandora • Pandora reconstruction performance • DUNE Far Detector • ProtoDUNE-SP • Summary Leigh Whitehead - University of Cambridge 2
Introduction to DUNE • DUNE is a next-generation neutrino oscillation experiment South Dakota Illinois • Wide physics programme • Neutrino oscillations: • Measurement of CP violation in the neutrino sector • Neutrino mass hierarchy • High precision measurement of oscillation parameters • BNV, supernova neutrinos, sterile neutrinos, etc Leigh Whitehead - University of Cambridge 3
Introduction to DUNE • DUNE is a next-generation neutrino oscillation experiment South Dakota Illinois • High power neutrino beam produced at Fermilab • Neutrino beam characterised by a suite of Near Detectors • Far Detectors measure the oscillated neutrino beam • Four 10kt (fiducial) liquid argon time projection chambers (LArTPCs) • Focus on the single-phase TPC design in this talk • Three views of each neutrino interaction from wire readout planes Leigh Whitehead - University of Cambridge 4
Single Phase LArTPC • Detector fully submerged in liquid argon /cm) (5 00V i c Field Anode tr Elec Ionising particle Drifting electrons produce signals ( ) in the blue and green induction wires and are collected by the purple wires Ionisation electrons Cathode drift in the electric Time field Wire Leigh Whitehead – CERN – ICHEP 2018 5
Introduction to Pandora • Pandora is a software development kit for automated event reconstruction • Originally for the ILC detectors • Now applied to a number of LArTPC neutrino experiments • Collaboration between a number of UK institutes • New members and collaborators welcome! • UK long-term Pandora plan for DUNE defined for 7 years Key references: Eur. Phys. J. C (2018) 78: 82 and Eur. Phys. J. C (2015) 75: 439 Leigh Whitehead - University of Cambridge 6
Multi-algorithm approach • Break down problem into many smaller steps, and gradually build up features in events • Pandora is a software framework to facilitate this multi- algorithm approach • Each step is incremental and aims to avoid mistakes (hard to undo!) • As the event develops employ more sophisticated algorithms • Over 100 algorithms and tools for LArTPCs • Use cluster-based pattern recognition where it works well • Machine (and deep) learning approaches to help make key decisions • Build physics and detector knowledge into the algorithms Leigh Whitehead - University of Cambridge 7
Reconstruction Steps I - III • Build up events in stages Simulated CC vu interaction in DUNE FD 1) 2) 3) Gap between volumes Gap between volumes Gap between volumes Gap between volumes Initial Refined clusters clusters Topological association algorithms w, wire position w, wire position w, wire position Candidate vertex positions x, drift position x, drift position x, drift position • Form basic clusters in 2D and then merge them together topologically • Detector geometry gives information such as gaps between volumes • Form candidate 3D vertices Leigh Whitehead - University of Cambridge 8
Reconstruction Steps IV - VI • Fully reconstruct in 3D Simulated CC vu interaction in DUNE FD 4) u:v:w 5) 6) Gap between volumes Gap between volumes Gap between volumes Initial Output 1:1:1 particles particles u u cluster Protected track Primary Daughter clusters shower shower particle Particle Primary merging track v Candidate algorithms Primary shower branches track v cluster Candidate shower seeds Primary w shower w, wire position w, wire position u, v → w predictions + Two small superimposed on w cluster primary tracks near vertex Parent x, drift position x, drift position x, drift position interaction vertex • Form tracks in 3D by matching between 2D views • Identify tracks and showers • Refine particle creation going back and forth between 2D and 3D • Build the events and hierarchy Leigh Whitehead - University of Cambridge 9
Far Detector performance • Particle reconstruction efficiencies DUNE Technical Design Report Volume II: https://arxiv.org/abs/2002.03005 Leigh Whitehead - University of Cambridge 10
Far Detector performance • Neutrino interaction vertex resolution DUNE Technical Design Report Volume II: https://arxiv.org/abs/2002.03005 Leigh Whitehead - University of Cambridge 11
ProtoDUNE-SP • ProtoDUNE-SP is a DUNE prototype detector • Located in a particle test-beam at CERN • Surface location - large cosmic flux • Central cathode divides the detector into two drift volumes See the ProtoDUNE talks and posters given at this conference for more information Leigh Whitehead - University of Cambridge 12
Pandora in ProtoDUNE-SP • Pandora is the primary reconstruction for ProtoDUNE-SP • The outputs are used for the large number of analyses Zoom of beam interaction • Suite of analysis tools to best use the Pandora outputs • Feedback from analysers Simulated test-beam interaction in ProtoDUNE DUNE Technical Design Report Volume II: https://arxiv.org/abs/2002.03005 Leigh Whitehead - University of Cambridge 13
Pandora in ProtoDUNE-SP • ProtoDUNE-SP cosmic data and MC • Stitch collinear tracks either side of the cathode or anode using an equal and opposite shift in the drift direction • Measure the arrival time in the TPC, t0, for this subset of tracks Left: DUNE Technical Design Report Volume II: https://arxiv.org/abs/2002.03005 Right: ProtoDUNE Performance Paper: https://arxiv.org/abs/2007.06722 Leigh Whitehead - University of Cambridge 14
New developments • There is a lot of on-going work to improve all aspects of the event reconstruction • Integration of deep learning techniques • Convolution Neural Network (CNN) for vertex finding • CNN for identifying track- and shower-like hits • Increase use of charge information • Matching clusters between the different readout planes • Feature finding in geometrically ambiguous cases • High energy muons • Important for calibration in the Far Detectors • Improve track finding through large delta-ray showers • Stitch tracks across multiple cathodes and anodes Leigh Whitehead - University of Cambridge 15
Summary • Pandora is performing well for DUNE and ProtoDUNE-SP • We are also supporting ProtoDUNE-DP • Many improvements planned to all steps of the reconstruction chain • Improvements will be benchmarked on DUNE FD simulation and, where possible, using ProtoDUNE data and simulation • Strong intention to support DUNE Near Detector reconstruction (LArTPC, GArTPC, scintillator trackers) Leigh Whitehead - University of Cambridge 16
Pandora Pandora is an open project and new contributors would be extremely welcome. We’d love to hear from you and we will always try to answer your questions. https://github.com/PandoraPFA https://pandorapfa.slack.com John Marshall Leigh Whitehead Andy Blake Andy Chappell Lorena Escudero Dom Brailsford Maria Brunetti Steve Green Ryan Cross Jhanzeb Ahmed Jack Anthony Mousam Rai Andy Smith Isobel Mawby Stefano Vergani Alex Moor ProtoDUNE-DP material from Etienne Chardonnet (Université Paris Diderot / Laboratoire APC) Pandora SDK: John Marshall, Mark Thomson ✦ Core LArTPC algs: Andy Blake, John Marshall Leigh Whitehead - University of Cambridge 17
You can also read
