Surrogate forward models for population inference on compact binary mergers

TL;DR

This paper introduces a neural network-based surrogate model for rapid inference of astrophysical and cosmological parameters from gravitational-wave observations of binary black hole mergers, enhancing analysis efficiency.

Contribution

It develops a fast, flexible surrogate model using neural networks to emulate binary population synthesis, enabling efficient inference of astrophysical and cosmological parameters.

Findings

Surrogate model accurately emulates population synthesis simulations.

Method constrains star formation rate and metallicity evolution.

Enables rapid analysis of gravitational-wave catalog data.

Abstract

Rapidly growing catalogs of compact binary mergers from advanced gravitational-wave detectors allow us to explore the astrophysics of massive stellar binaries. Merger observations can constrain the uncertain parameters that describe the underlying processes in the evolution of stars and binary systems in population models. In this paper, we demonstrate that binary black hole populations - namely, detection rates, chirp masses, and redshifts - can be used to measure cosmological parameters describing the redshift-dependent star formation rate and metallicity distribution. We present a method that uses artificial neural networks to emulate binary population synthesis computer models, and construct a fast, flexible, parallelisable surrogate model that we use for inference.