Nonparametric analysis of correlations in the binary black hole population with LIGO-Virgo-KAGRA data

Formation channels of merging compact binaries imprint themselves on the distributions and correlations of their source parameters. But current understanding of this population observed in gravitational waves is hindered by simplified parametric models. We overcome these limitations using PixelPop Heinzel et al. (2024)-our multidimensional Bayesian nonparametric population model. We analyze data from the first three LIGO-Virgo-KAGRA observing runs and make high resolution, minimally modeled measurements of the pairwise distributions of binary black hole masses, redshifts, and spins. There is no evidence that the mass spectrum evolves over redshift and we show that such measurements are fundamentally limited by the detector horizon. We find support for correlations of the spin distribution with binary mass ratio and redshift, but at reduced significance compared to overly constraining parametric models. Confident data-driven conclusions about population-level correlations using very flexible models like PixelPop will require more informative gravitational-wave catalogs.