TL;DR
This paper details how the Pandora software framework was adapted to reconstruct particle interactions in the ProtoDUNE-SP liquid argon time projection chamber. It highlights the challenges of complex events with background noise and demonstrates that Pandora achieves high reconstruction and identification efficiencies for test-beam particles, crucial for future neutrino experiments like DUNE.
What Problem Does This Paper Address?
Reconstructing particle interactions in large liquid argon time projection chambers (LArTPCs) like ProtoDUNE-SP is a significant challenge. These detectors are exposed to a high flux of cosmic rays and other background particles, making it difficult to isolate and accurately reconstruct the desired neutrino or test-beam particle signals. For future neutrino experiments such as DUNE, precise reconstruction of particle trajectories and interactions is essential for accurate measurements of fundamental physics processes, like neutrino cross-sections. This paper addresses the need for a robust and efficient reconstruction software that can handle the complexity of these detectors and their data.
Methods
The core of the paper focuses on the Pandora Software Development Kit (SDK) and its application to ProtoDUNE-SP. Pandora employs a multi-algorithm, pattern-recognition approach to reconstruct particle events. Key aspects of their method include:
- Hierarchical Reconstruction: Pandora builds a hierarchy of reconstructed particles, starting from individual hits and progressively grouping them into clusters, then particles, and finally complex interaction hierarchies.
- Multi-View Reconstruction: It utilizes three 2D views (u, v, w wire planes) from the LArTPC to reconstruct particle trajectories in 3D.
- Algorithm Chains: Two main algorithm chains were developed and adapted:
- PandoraCosmic: Optimized for reconstructing cosmic-ray muons, which are abundant and can traverse the detector. This chain includes a "stitching" algorithm to handle ambiguities arising from multiple drift volumes and space charge effects.
- PandoraTestBeam: Specifically designed for reconstructing test-beam particles, focusing on identifying interaction vertices and the resulting particle showers.
- Consolidated Reconstruction: A novel approach that combines the outputs of PandoraCosmic and PandoraTestBeam, using a machine learning classifier (BDT) to select the most appropriate reconstruction for each event slice.
- Space Charge Effect (SCE) Mitigation: The reconstruction accounts for distortions caused by the accumulation of positive ions from cosmic-ray muons, which can affect electric fields and particle trajectories.
Datasets
- Simulated Data:
- GEANT4 simulations of the beamline and particle interactions.
- CORSIKA simulations for cosmic rays.
- Data-driven simulation of space charge effects.
- Specialized simulation samples to study systematic uncertainties:
- SCE-off sample (no space charge simulation).
- Reduced beam halo sample (15% reduction).
- Experimental Data:
- Data collected from the ProtoDUNE-SP detector at CERN during August-December 2018, using a charged particle test beam.
- Data was collected at specific beam momenta: 1, 2, 3, 6, and 7 GeV/c.
Key Insights and Ideas
- Pandora's Adaptability: The Pandora framework, originally developed for other experiments, proved highly adaptable to the unique challenges of LArTPCs like ProtoDUNE-SP, particularly its multi-drift-volume geometry.
- The "Stitching" Algorithm: The ability to "stitch" reconstructed tracks across different drift volumes is crucial for accurately determining the timing of cosmic-ray muons and mitigating ambiguities.
- Consolidated Reconstruction for Clarity: Combining outputs from specialized chains (Cosmic vs. Test-Beam) via a BDT is an effective strategy to achieve a clear and unambiguous event reconstruction, especially in complex scenarios.
- Space Charge is a Real Factor: The paper emphasizes the impact of space charge effects on reconstruction accuracy and details how their reconstruction accounts for these distortions.
- Simulation vs. Data Agreement: The close agreement between simulated and experimental data efficiencies (within 5%) validates both the simulation's accuracy and Pandora's performance.
Key Results
- High Reconstruction Efficiency: Simulated reconstruction and identification efficiency for triggered test-beam particles exceeds 80% for most particle types and momenta.
- Specific Efficiencies:
- 1 GeV/c charged pions: 86.1 ± 0.6%
- 1 GeV/c protons: 84.1 ± 0.6%
- Cosmic Ray Tagging: Approximately 62.4% of cosmic-ray muon candidates were successfully tagged using the defined criteria.
- Test-Beam Data Agreement: Measured efficiencies for test-beam data are within 5% of simulated predictions.
- Position Resolution: For 1 GeV/c particles, the 3D end-point position was reconstructed within 2 cm of the true value for 68% of events.
Limitations and Future Work
- Limitations:
- The paper notes that it's not trivial to extrapolate ProtoDUNE-SP performance to the DUNE far detector due to ProtoDUNE-SP's higher detector occupancy.
- Some discrepancies between data and simulation are observed, potentially due to:
- Underestimation of space charge effects in simulation.
- Differences in the ratio of muons to pions between data and simulation.
- The presence of a malfunctioning electron diverter affecting beam particle tracks.
- The analysis of charged kaons and muons was limited due to data collection constraints.
- Future Work:
- Development of new algorithms and incorporation of deep-learning techniques for improved vertex finding, event slicing, and hit classification.
- These advancements will be tested on ProtoDUNE-SP simulation and data, and will be crucial for the DUNE far detector.
- A similar upgraded detector, ProtoDUNE-HD, is under construction and will provide further testing grounds.
Summary updated