Neutron stars are cosmic laboratories where matter exists in states impossible to replicate on Earth. Understanding their internal structure requires solving the Tolman-Oppenheimer-Volkoff (TOV) equations and inferring nuclear equations of state from observational data - a computationally intensive challenge that has limited our ability to unlock these stellar secrets in real-time.
JESTER transforms this bottleneck with JAX-accelerated computing, bringing GPU power to neutron star physics. The package supports multiple EOS parametrizations including metamodels, speed-of-sound extrapolations, and spectral expansions from Lindblom 2010. It integrates cutting-edge Bayesian samplers - Sequential Monte Carlo, nested sampling via BlackJAX, and normalizing flow-enhanced MCMC through flowMC - enabling researchers to efficiently explore parameter spaces that were previously computationally prohibitive.
This tool arrives at a golden age of multimessenger astronomy, where LIGO gravitational wave detections and NICER X-ray observations provide unprecedented constraints on neutron star properties. JESTER’s GPU acceleration makes it practical to perform real-time inference on these rich datasets, helping astronomers decode the fundamental physics governing matter at nuclear densities and contributing to our understanding of stellar evolution and cosmic phenomena.
⭐ Stars: 10
💻 Language: Python
🔗 Repository: nuclear-multimessenger-astronomy/jester