Skip to content

KM3NeT/KM3NeT-ARCA-irf

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

License

KM3NeT/ARCA230 Instrument response functions

This repository contains the instrument response functions of the full KM3NeT/ARCA230 detector. The IRFs are accompanied by a set of scripts to interact with the IRFs and to perform an example cut-and-count analysis to calculate the sensitivity and discovery potential to a neutrino point source.

N.B.: The resulting sensitivity and discovery potential is worse than presented in the paper due to:

  • The cut-and-count method only looks at the track (or shower) channel instead of combining both,
  • This analysis only includes signal from $\nu_\mu$ and $\bar{\nu}_\mu$ CC events selected as track and $\nu_e$ and $\bar{\nu}_e$ CC events selected as shower, instead of all flavours and interactions,
  • The paper uses a more sophisticated method than presented here. The paper uses a binned likelihood method and throws pseudo experiments to determine the sensitivity, while in this example we use Poisson statistics for a simple counting experiment.

Content

  • data/: Instrument Response Functions (IRFs) for the KM3NeT/ARCA230 detector.
  • analysis/: Jupyter notebooks with example plots and analysis
  • src/arca230/:
    • flux.py: Class that represents a single power law neutrino point source flux.
    • aeff.py: Class that loads the effective area and calculates event rates using a point source flux.
    • psf.py: Class that loads the point spread function and calculates probabilities to reconstruct events with a specified search cone size.
    • energyresponse.py: Class that loads the energy response and convolves true neutrino energies with the energy response of the detector.
    • background.py: Class that calculates expected background rates at different positions in the sky.

Installation

Download

The content of the repository is downloaded via git:

git clone git@git.km3net.de:open-data/public-candidates/open-point-source-search.git

Creating the environment

Using venv

Create a virtual environment

python -m venv my_venv

Source the virtual environment

source my_venv/bin/activate

Install

Enter the dowloaded repository

cd open-point-source-search

and install the requirements

pip install -e .

Running Jupyter notebooks

In order to run the notebooks, you need to install Jupyter, by using pip install jupyter or following the instructions at the Juypter website.

Running the Jupyter kernel

From within your virtual environment, create a Jupyter kernel and launch your notebook:

python -m ipykernel install --user --name=km3net_ps
jupyter-notebook

For zsh shell, you need to execute these lines first before installation of the kernel

conda install -c conda-forge notebook
conda install -c conda-forge nb_conda_kernels

You can then execute the notebooks in your browser following the URL in the stdout.