ANTARES Processing Pipeline

Overview

Pipeline to convert ANTARES event files (AntDSTs) into a format compatible with the Swim analysis framework. The pipeline processes and merges files, calculates oscillation weights, and integrates reconstructed event data from the NNFit algorithm. It has been tested for both Monte Carlo simulation and recorded events.

Input Files

  • AntDSTs: ROOT-based summary files containing:
    • Run info & quality parameters
    • MC truth parameters (energy, direction, vertex, weights) -> Only available in MC
    • Reconstructed parameters from algorithms (AAFit, BBFit, GridFit, Showerdusj, Showertantra)

Pipeline Steps

1. File Extraction

  • Extract relevant branches from AntDSTs into standard ROOT TTrees.
  • Includes:
    • Data quality (DQ) tree: run ID, duration, environmental conditions
    • MC truth tree: true energy, direction, vertex, weights
    • Reconstruction trees: algorithm-specific reconstructed parameters

2. Intermediate Merging

  • Merge files by event type and run number to reduce file count and improve processing efficiency.
  • 15 event categories (e.g., \(\nu_{e}^{CC}\), \(\nu_{\mu}^{CC}\), \(\nu_{\tau}^{CCmu}\), \(\mu\), etc.)

3. Correction & Weight Application

  • Apply corrections to reconstruction branches (fix misassigned values).
  • Apply weight corrections:
    • 20% reduction for runs < 30412 (early detector overestimation)
    • Data/MC corrections based on year and event type
    • Compute weight_one_year for muons

4. Oscillation Weight Calculation

  • Use OscProb library to compute oscillated weights for neutrino events.
  • Weights calculated as: \(w_{\text{osc}} = \frac{w_2}{ngen} \cdot (\Phi_{\nu_e} P_{\nu_e \to \nu_i} + \Phi_{\nu_\mu} P_{\nu_\mu \to \nu_i}) \cdot \text{run_duration}\)
  • Uses Honda atmospheric flux model and PREM Earth model.

5. Second Merging (NNFit Integration)

  • Merge files into groups matching NNFit HDF5 file structure.
  • Classes: Showers, NuMuCC, Taus, Muons

6. NNFit Branch Addition

  • Merge NNFit reconstructed parameters into the main ROOT file using:
    • Run ID
    • Frame index
    • Event ID
This image can also have a caption. It's like magic.

7. Final Merging & Preparation

  • Apply selection cuts (e.g., low-energy up-going, NNFit quality cuts).
  • Add “MC-truth” branches for perfect reconstruction reference.
  • Include detector dimensions (cylinder: \(z_{\text{min}}, z_{\text{max}}, r\)) for Swim compatibility.

Output

  • Single ROOT file containing:
    • True and reconstructed parameters from all algorithms
    • Oscillation and correction weights
    • Classifier and selection cut flags
    • Compatible with Swim for sensitivity analysis

Key Features

  • Handles over 300,000 files efficiently
  • Supports both neutrino and muon events
  • Integrates modern ML-based reconstruction (NNFit)
  • Modular and parallelizable via cluster computing