Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Optimal Transport for $e/π^0$ Particle Classification in LArTPC Neutrino Experiments

David Caratelli, Nathaniel Craig, Chuyue Fang, Jessica N. Howard

Abstract

The efficient classification of electromagnetic activity from $π^0$ and electrons remains an open problem in the reconstruction of neutrino interactions in Liquid Argon Time Projection Chamber (LArTPC) detectors. We address this problem using the mathematical framework of Optimal Transport (OT), which has been successfully employed for event classification in other HEP contexts and is ideally suited to the high-resolution calorimetry of LArTPCs. Using a publicly available simulated dataset from the MicroBooNE collaboration, we show that OT methods achieve state-of-the-art reconstruction performance in $e/π^0$ classification. The success of this first application indicates the broader promise of OT methods for LArTPC-based neutrino experiments.

Optimal Transport for $e/π^0$ Particle Classification in LArTPC Neutrino Experiments

Abstract

The efficient classification of electromagnetic activity from and electrons remains an open problem in the reconstruction of neutrino interactions in Liquid Argon Time Projection Chamber (LArTPC) detectors. We address this problem using the mathematical framework of Optimal Transport (OT), which has been successfully employed for event classification in other HEP contexts and is ideally suited to the high-resolution calorimetry of LArTPCs. Using a publicly available simulated dataset from the MicroBooNE collaboration, we show that OT methods achieve state-of-the-art reconstruction performance in classification. The success of this first application indicates the broader promise of OT methods for LArTPC-based neutrino experiments.

Paper Structure

This paper contains 33 sections, 2 equations, 13 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Liquid Argon Time Projection Chamber Neutrino Detectors
  3. Physics Landscape with LArTPC Detectors
  4. Introduction to LArTPC detectors
  5. Detector Simulation and Low-Level Reconstruction
  6. MicroBooNE Open Dataset
  7. Detector Simulation and Calibrations
  8. Low-Level Reconstruction
  9. Current Status of LArTPC Pattern Recognition
  10. $e/\pi^0$ Particle Classification
  11. Current Status of LArTPC Reconstruction for $e/\pi^0$
  12. Optimal Transport Distances for LArTPC Detectors
  13. Review of Discrete OT Distances
  14. $e/\pi^0$ Classification with OT
  15. Data Pre-processing
  16. ...and 18 more sections

Figures (13)

  • Figure 1: Example event display of a neutrino interaction collected with one wire plane of the MicroBooNE LArTPC detector. Four photon EM showers, likely from the decay of two $\pi^0$ mesons, can be seen branching out from the interaction vertex. The x-axis represents the wire position while the y-axis indicates the time taken for the electrons to drift to the wire plane. The color scale in the event display shows the measured charge deposit.
  • Figure 2: True energy distribution of $e$ and $\pi^0$ particles in the MicroBooNE open datasets used for this work.
  • Figure 3: Example event displays for $\pi^0$ (a) and $\mathrm{e}^-$ (b).
  • Figure 4: Events from Fig. \ref{['fig:EVD-Events']} after 3D reconstruction and alignment. The size of the dots represent the relative amount of energy deposited associated with each reconstructed 3D spacepoint. (a) depicts a $\pi^0$ event and shows how the principal axis alignment based on the largest cluster in the event aligns the image with one of the two photon showers, keeping the second shower at an angle with respect to the $\hat{z}$ direction. (b) depicts an electron shower.
  • Figure 5: Example transport plans for optimal transport distances computed between two 3D event reconstructions. (a) transport plan between $\pi^0$ event (red) vs. ${e}^-$ event (blue); (b) transport plan between the same ${e}^-$ event (blue) and another $\mathrm{e}^-$ event (turquoise).
  • ...and 8 more figures