Measuring correlations in realistic astrophysical populations of gravitational-wave sources with PixelPop

The origins of merging compact binaries observed by the LIGO-Virgo-KAGRA gravitational-wave (GW) detectors remain highly uncertain, with disparate astrophysical processes possibly contributing to the overall merger rate. Such formation channels result in nontrivial correlations in the underlying distribution of source parameters. Current understanding of these sources relies heavily on simplified parametric models that make strong assumptions about the population. Recent advances have introduced more flexible methods like PixelPop – a high-resolution Bayesian nonparametric population model designed to infer joint distributions and parameter correlations with minimal assumptions. In this work, using realistic population synthesis simulations from different formation channels, we analyze two-, three-, and four-dimensional correlations in the masses, spins and redshift for simulated black-hole mergers observable by LIGO with current sensitivity. We show that astrophysical insights are limited and even significantly biased when higher dimensional correlations are neglected. However, using PixelPop allows us to correctly measure the astrophysical merger rate in all source parameters without modeling assumptions. Given our uncertainty in the formation of GW sources, robust extraction of astrophysics from GW catalogs requires that we allow for potential nontrivial high dimensional correlations with models like PixelPop.