The detection of multimessenger astrophysical signals is one of the most exciting and significant new areas in physics. However, extracting meaningful physical understanding from observations depends fundamentally on accurately modeling and simulating such complex multiphysics, multiscale systems. For example, neutrinos play a key role in neutron star mergers, but the extreme complexity and scale of this class of problem force the use of severe, and often inadequate, approximations. This project directly addresses these deficiencies with a novel approach, which will be open-source and incorporated into the widely-used FLASH radiation hydrodynamics code, thus benefiting the broader computational astrophysics community. The research includes significant mentoring of graduate students and postdoctoral researchers, who will develop skills that are in high demand.
The work will advance the state-of-the-art in modeling neutrino transport in high-accuracy simulations of binary neutron star mergers and core-collapse supernovae, developing surrogate models for the kinetic equations by using machine learning (ML) and reduced order modeling (ROM), to achieve scalable multi-scale transport simulations. The framework, dubbed Surrogate Transport Adaptive Model Procedure (STAMP), replaces algebraic closure with adaptive computed methods from a surrogate model, and retains the highly-scalable nature of moment methods while recovering the accuracy of multi-angle approaches. This project advances the goals of the NSF Windows on the Universe Big Idea.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
