1. Neural Simulation-Based Inference of the Neutron Star Equation of State directly from Telescope Spectra
- Author
-
Brandes, Len, Modi, Chirag, Ghosh, Aishik, Farrell, Delaney, Lindblom, Lee, Heinrich, Lukas, Steiner, Andrew W, Weber, Fridolin, and Whiteson, Daniel
- Subjects
Astronomical Sciences ,Physical Sciences ,astro-ph.HE ,astro-ph.IM ,gr-qc ,hep-ph ,nucl-th - Abstract
Neutron stars provide a unique opportunity to study strongly interactingmatter under extreme density conditions. The intricacies of matter insideneutron stars and their equation of state are not directly visible, butdetermine bulk properties, such as mass and radius, which affect the star'sthermal X-ray emissions. However, the telescope spectra of these emissions arealso affected by the stellar distance, hydrogen column, and effective surfacetemperature, which are not always well-constrained. Uncertainties on thesenuisance parameters must be accounted for when making a robust estimation ofthe equation of state. In this study, we develop a novel methodology that, forthe first time, can infer the full posterior distribution of both the equationof state and nuisance parameters directly from telescope observations. Thismethod relies on the use of neural likelihood estimation, in which normalizingflows use samples of simulated telescope data to learn the likelihood of theneutron star spectra as a function of these parameters, coupled withHamiltonian Monte Carlo methods to efficiently sample from the correspondingposterior distribution. Our approach surpasses the accuracy of previousmethods, improves the interpretability of the results by providing access tothe full posterior distribution, and naturally scales to a growing number ofneutron star observations expected in the coming years.
- Published
- 2024