toise

toise is a tool for estimating the sensitivity of natural-medium neutrino detectors such as IceCube-Gen2 to sources of high-energy astrophysical neutrinos. It uses parameterizations of a detector's fiducial area or volume, selection efficiency, energy resolution, angular resolution, and event classification efficiency to convert (surface) neutrino fluxes into mean event rates in bins of observable space. These are then used to estimate statistical quantities of interest, e.g. the median sensitivity to some flux (i.e. 90% upper limit assuming the true flux is zero) or the median discovery potential (i.e. the flux level at which the null hypothesis would be rejected at 5 sigma in 50% of realizations).

A toise is also an archaic French unit of length, area, or volume, depending on context. toise may also emit projected sensitivities in different units depending on context.

Installation

The way to install toise and its depedencies is with micromamba¹. If you do not already have micromamba on your system,

"${SHELL}" <(curl -L micro.mamba.pm/install.sh)

or, os macOS with Homebrew:

brew install micromamba

Next, obtain toise:

git clone [email protected]:icecube/toise.git

Then, from the toise source directory, create a new environment:

cd toise
micromamba create -n toise --file conda-lock.yml
micromamba run -n toise pip install -e .

If you have a Jupyter notebook installation from another conda environment, you should now be able to open this notebook in Jupyter and select the "Python [conda env:miniconda3-toise]" kernel.

Otherwise, activate the environment with micromamba activate toise. This should leave you with a prompt that looks like

(toise) [jakob@TheInfoSphere3:~]$.

If you'd like to install jupyter at this stage, you can do:

micromamba install --channel conda-forge notebook

Adding your detector to toise

toise uses parameterized response functions to calculate expected sensitivities to scenarios of astrophysical neutrino production. It comes with parameterizations for a fictive under-ice optical detector as well as a radio detector. To use toise with your own detector design, you will have to provide parameterizations its performance.

Optical detector

For an optical detector, you will need to provide:

1. The instrumented volume of the detector.
2. For muons that enter the instrumented volume from the outside:
   1. Selection efficiency, as a function of muon energy at the boundary of the instrumented volume and zenith angle.
   2. Energy resolution, in the form of a transfer matrix from muon energy at the boundary of the instrumented volume to energy observable.
   3. Angular resolution, in the form of a cumulative distribution of opening angle between the true and reconstructed muon direction, as a function of muon energy at the boundary of the instrumented volume and zenith angle.
3. For neutrino interactions inside the instrumented volume:
   1. Selection efficiency, as a function of deposited energy and zenith angle.
   2. Energy resolution, in the form of a transfer matrix from deposited energy to energy observable.
   3. Angular resolution, in the form of a cumulative distribution of opening angle between the true and reconstructed neutrino direction, as a function of deposited energy and zenith angle.
   4. Flavor signature classification efficiency, in terms of an energy-dependent transfer matrix from interaction type to flavor signature (cascade, double cascade, starting track).

Radio detector

For a radio detector, you will need to provide:

1. The (zenith and neutrino energy dependent) effective volumes for neutrinos at trigger level
2. (Optionally) The PDF of triggered shower energies for given neutrino energy. Otherwise the default transfer matrices of provided within toise will be used.

Default parametrization functions for analysis efficiency, energy resolution and angular point spread are available within the framework. The provided function parameters can be steered by the configuration .yaml files. To find appropriate parameters describing the detector to be studied, producing the following distributions is sufficient:

3. The efficiency with which triggered events can be reconstructed and analysed as a function of shower energy.
4. The (1D) energy resolution in terms of log10(E_rec/E_shower).
5. The angular resolution point spread function. If using the default double-Gaussian implementation, the fraction of well reconstrucable events and the fraction of poorly reconstructable events with corresponding sigma. These angular resolution quantities may be provided energy dependently.

Footnotes

1. micromamba is a fast, small, [mostly] drop-in replacement for the conda package manager. ↩