Removing degeneracy and multimodality in gravitational wave source parameters

TL;DR

This paper introduces new coordinate systems and methods to reduce degeneracy and multimodality in gravitational wave source parameter estimation, improving efficiency and robustness of inference.

Contribution

It develops coordinate transformations and algorithms to disentangle parameters and address multimodality caused by symmetries, with an open-source Python package for implementation.

Findings

Coordinates effectively remove correlations among parameters.

Identified symmetries as main cause of multimodality in posteriors.

Introduced a well-measured spin azimuth potentially indicating orbital precession.

Abstract

Quasicircular binary black hole mergers are described by 15 parameters, of which gravitational wave observations can typically constrain only $\sim 10$ independent combinations to varying degree. In this work, we devise coordinates that remove correlations, and disentangle well- and poorly-measured quantities. Additionally, we identify approximate discrete symmetries in the posterior as the primary cause of multimodality, and design a method to tackle this type of multimodality. The resulting posteriors have little structure and can be sampled efficiently and robustly. We provide a Python package for parameter estimation, cogwheel, that implements these methods together with other algorithms for accelerating the inference process. One of the coordinates we introduce is a spin azimuth that is measured remarkably well in several events. We suggest this might be a sensitive indicator of orbital precession, and we anticipate that it will shed light on the occurrence of spin-orbit misalignment in nature.