EMRI Search

This repository implements a search pipeline for detecting Extreme Mass Ratio Inspirals (EMRIs) in simulated LISA data. Given input LISA time-series, the pipeline returns the best-matching instantaneous frequency evolution (track) found in the data. The code accompanies the related paper: Ab uno disce omnes: Single-harmonic search for extreme mass-ratio inspirals. Intermediate search results can be found here. If you are interested only in which sources we were able to detect check these interactive visualizations.

For a quick, simplified EMRI search, open and run QuickStartEMRIsearch.ipynb. The notebook walks through the pipeline and produces example outputs, including:

• Injected waveform and three harmonics with their corresponding detection statistic values.

  

• Top panel: the injected EMRI signal with noise. Bottom panel: the recovered signal after masking the best-fit track of the dominant harmonic.

  

• Visualizations of the optimization process and the recovered track evolution.

Table of Contents

• Installation
• Quick Start
• Repository Structure
• Core Utilities
• Citation
• Contact

Installation

Create a conda environment

Recommended minimal environment:

conda create -n emri -y python=3.12
conda activate emri
pip install -e .

with the flag -e for development.

There is one package that needs to be installed from source fastlisaresponse. On macOS arm64 a working build sequence used by the author:

git clone https://github.com/mikekatz04/lisa-on-gpu.git
cd lisa-on-gpu
git checkout v1.2.1a0

conda install conda-forge::clang_osx-arm64 conda-forge::clangxx_osx-arm64 -y
export CC=$(which clang)
export CXX=$(which clang++)
export CFLAGS="-march=native"
export CXXFLAGS="-march=native"

pip install .
python -m unittest discover

Test most of the functions of the package with:

python -m unittest tests.test_demo

Quick Start

Open QuickStartEMRIsearch.ipynb and run the notebook. It guides you through:

• generating EMRI test signals,
• computing Short Fourier Transforms (SFTs),
• computing the detection statistic,
• and running example searches that reproduce the figures above.

Repository Structure

High-level overview of the modules's key files and their purposes. Each script can be run to generate a demonstration plot showing typical usage of the functions it defines.

Core Utilities

Data analysis and waveform helpers:

• da_utils.py — SFT computation, frequency-domain inner products, noise generation, SNR calculation.
• emri_utils.py — EMRI waveform and trajectory helpers
• search_utils.py — detection statistics and search routines
• draw_population.py — draw synthetic EMRI populations and parameter transforms.

JAX / GPU-Accelerated Utilities

• jax_utils.py — JAX implementations of core routines for GPU/TPU acceleration (PSDs, detection statistics).
• jax_de_utils.py — differential-evolution optimization implemented with JAX for fast population-based searches.

Citation

If you use this code, please cite the paper:

@ARTICLE{2026PhRvD.113b4061S,
       author = {{Speri}, Lorenzo and {Tenorio}, Rodrigo and {Chapman-Bird}, Christian and {Gerosa}, Davide},
        title = "{Single-harmonic search for extreme mass-ratio inspirals}",
      journal = {\prd},
     keywords = {General relativity, alternative theories of gravity, General Relativity and Quantum Cosmology, High Energy Astrophysical Phenomena, Instrumentation and Methods for Astrophysics},
         year = 2026,
        month = jan,
       volume = {113},
       number = {2},
          eid = {024061},
        pages = {024061},
          doi = {10.1103/dh3j-ksfl},
archivePrefix = {arXiv},
       eprint = {2510.20891},
 primaryClass = {gr-qc},
       adsurl = {https://ui.adsabs.harvard.edu/abs/2026PhRvD.113b4061S},
      adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}

Contact

For questions or issues open an issue on the GitHub repository.