A Hybrid Neural Network Emulator for M1 Neutrino Reaction Rates in Binary Neutron Star Merger Simulations

Abstract

We present a neural network emulator for the bns_nurates library, which computes neutrino reaction rates used in binary neutron star merger simulations. Following a hybrid strategy, we emulate only the computationally dominant two-dimensional reactions---pair processes, bremsstrahlung, and inelastic neutrino-electron scattering---while evaluating the inexpensive one-dimensional processes still with bns_nurates. We find that the emulator reproduces the opacities to within a few percent across the relevant thermodynamic conditions and accelerates their evaluation by roughly an order of magnitude. When deployed in a global Athenak binary neutron star merger simulation, it recovers the remnant structure, neutrino luminosities, and ejecta composition of a reference run using the full bns_nurates library, at approximately half the overall cost

Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.