Continuous-wave search sensitivity simulator (cows3)

A Python package to estimate the sensitivity of general continuous gravitational-wave searches.

The method should be equivalent to the semi-analytical approach derived in Dreissigacker, Prix, Wette (2018) and implemented in Octapps, but here we implement it in Python to make it more convenient to use.

How to install

cows3 is available under PyPI:

pip install cows3

Citing this work

If cows3 was useful to your research, we would appreciate if you cited Mirasola & Tenorio (2024) where this implementation was first presented:

@article{Mirasola:2024lcq,
    author = "Mirasola, Lorenzo and Tenorio, Rodrigo",
    title = "{Toward a computationally efficient follow-up pipeline for blind continuous gravitational-wave searches}",
    eprint = "2405.18934",
    archivePrefix = "arXiv",
    primaryClass = "gr-qc",
    reportNumber = "LIGO-P2400221",
    doi = "10.1103/PhysRevD.110.124049",
    journal = "Phys. Rev. D",
    volume = "110",
    number = "12",
    pages = "124049",
    year = "2024"
}

as well as a Zenodo release of this software.

For the semi-analytical sensitivity estimation method you should also cite Wette (2012) and Dreissigacker, Prix, Wette (2018). Also, this package makes extensive use of SWIG bindings, so please cite Wette (2021) as well.

Quick example

The examples/ directory contains a runnable script that estimates CW sensitivity for three detectors: H1, L1, and a custom PB_10k detector.

pip install cows3[examples]  # adds matplotlib
python examples/two_detector_different_psd.py

The script draws an isotropic population of CW sources, computes the unit-depth SNR² distribution, and evaluates the false-dismissal probability at selected sensitivity depths (Sec. II.3 of arXiv:2405.18934). It also saves a histogram of the SNR² distribution to result.pdf.

To experiment with the setup, edit the detector noise levels, observation time, and sensitivity depths directly in the script.

Note: This example concerns continuous-wave F-statistic sensitivity. It does not implement stochastic-background cross-correlation or overlap-reduction-function analyses.

Authors

• Rodrigo Tenorio
• Lorenzo Mirasola