University of Nottingham
Mathematics

From Earth, the distant stars appear eternal and isolated. On the contrary, stars have finite lifetimes and the majority accompany one or more stellar partners, bound into orbits by their mutual gravitational attraction. More massive stars are more likely to exist in these multiple systems but they also lead shorter lives. If massive enough, a dying star explodes in a magnificent supernova and leaves behind a black-hole remnant.

Given the abundance of stars in the universe, it should be littered with pairs of black holes orbiting each other. Their strong gravitational forces drive an inspiral toward a cataclysmic collision: a black-hole merger. These mergers emit no light but are unimaginably energetic, vibrating spacetime itself like ripples on water moving at the speed of light across the cosmos. Gravitational waves, as they are known, were first directly detected in 2015 — a century after Einstein predicted them — from two black holes 30 times heavier than the Sun that merged one billion years ago; this discovery was awarded the 2017 Nobel Prize in Physics. About 100 more signals have been observed since.

Despite this, the past lives of black-hole binaries remain poorly understood due to theoretical uncertainties in how they form. As binary systems are fundamental building blocks of the universe with stars producing all the chemical elements common in our daily lives, understanding the very end of theirs is crucial.

I propose a new paradigm to probe the formation of merging binary black holes. Current widespread analyses that use simplistic models for the population of sources will be replaced with a robust data-analysis framework that directly quantifies the match between state-of-the-art astrophysical predictions and real observations. I will train neural networks as efficient surrogates for population simulations that are too computationally costly to use directly for data analysis, improving previous models by using Bayesian deep learning to carefully account for training uncertainties. By constraining the input astrophysical parameters consistent with the catalogue of gravitational-wave observations using hierarchical inference, my work will revolutionize our understanding of how massive stars evolve in binaries, form black holes, and are driven to merge.