Revealing massive black hole astrophysics: The potential of hierarchical inference with extreme mass-ratio inspiral observations (data release)

This repository contains the data associated with Singh, Chapman-Bird, Veitch, and Berry (2026). We present a population analysis of extreme mass-ratio inspirals (EMRIs) observable by the LISA mission. Hierarchical Bayesian inference is performed using single-component and mixed population models for EMRIs, focusing on the primary black hole mass, secondary compact object mass, spin, and eccentricity parameters. These population models are designed to capture the diverse formation channels for EMRIs. For each model, we infer population hyperparameters using mock EMRI catalogs generated under consistent assumptions.

The configuration files and inference outputs provided here can be used together with poplar and the code in this GitHub repository.

Contents

The release includes data sufficient to reproduce all population-level results and figures in the paper.

• Configuration files (*.prior): Located at the top level of each population folder (pop_A, pop_B, pop_MIX/*, and DIFF_pop_INFERENCE/pop_MIX). These files specify the hyperparameter priors for the EMRI population models, including distributions for:
  • Primary (MBH) mass
  • Secondary compact object mass
  • Spin
  • Eccentricity
• Inference outputs: Each population folder contains subdirectories for different catalogue sizes (1E2_events, 1E3_events, 1E4_events). Within each, the inference folder includes:
  • config.json (pipeline configuration)
  • result.hdf5 (posterior samples of population hyperparameters) -- These result.hdf5 files are the primary data products underlying all population-level figures and quantitative results presented in the paper
  • state.png, trace.png (diagnostic plots)
• Plotting and Analysis Notebooks: A Jupyter notebook paper_plots.ipynb that loads the result files and plots them.