Speaker
Description
Binary neutron star (BNS) mergers are among the most information-rich events in modern astrophysics, producing a unique combination of multi-messenger signals including gravitational waves, short gamma-ray bursts, and kilonovae. These signals probe some of the most fundamental open questions in physics: the neutron star equation of state, r-process nucleosynthesis, and the origin of relativistic jets. Numerical simulations using General-Relativistic Magnetohydrodynamics (GRMHD) are essential to bridge the gap between observation and theory, yet their computational demands have historically limited their scope and resolution.
Modern high-performance computing clusters are increasingly GPU-powered, offering computational speedups of 10–50× over traditional CPU-based approaches. In this work, we present the first self-consistent BNS merger simulation within the GPU-accelerated GRMHD code GRaM-X, built on the Einstein Toolkit framework. The key missing ingredient was a compatible initial data solver; we ported the FUKA spectral initial data library into GRaM-X, enabling physically consistent starting conditions for the binary system. This required significant code restructuring, including GPU-compatible memory allocation, temperature initialization routines, and atmosphere handling tailored to BNS environments.
We demonstrate a stable 3D dynamical-spacetime GRMHD simulation of two 1.5 M☉ neutron stars through inspiral, tidal deformation, merger, and prompt collapse to a black hole. This work establishes a production-level BNS merger code using GRaM-X, opening the door to future studies of magnetic field amplification, neutrino-driven outflows, and multi-messenger signal modeling at significantly reduced computational cost.
| Talk category | NOVA Network 3 |
|---|
