Solutions to the Core-Collapse Supernova Problem from Theory

Adam Burrows, Princeton University

Usage Details

Core-collapse supernovae dramatically announce the death of massive stars and the birth of neutron stars and black holes. During this violent process, a combination of high-density nuclear physics, multi-dimensional hydrodynamics, radiation transport, and neutrino physics determines whether and how the star explodes. However, the precise mechanism of explosion has not been unambiguously pinned down and this fifty-year- old conundrum is one of the central remaining unsolved problems in theoretical astrophysics.

Recently, using our new state-of-the-art supernova code FORNAX , we simulated a large fraction of a planned suite of non-rotating three-dimensional (3D) supernova models. This suite included a set of runs of the collapse, bounce, and explosion (most often) of 9-, 10-, 11-, 12-, 13-, 14- , 15-, 16-, 19-, 25-, and 60-M progenitor massive stars and is the largest such 3D investigation in the history of core-collapse theory. All but the 11-M and 16-M models were performed on Blue Waters. 

Models that explode do so via the delayed neutrino heating mechanism with the default physics (neutrino-matter couplings, equation of state, approximate general relativity, radiation redshifts). This breakthrough was enabled by the fast HPC computational environments now available and by the novel code architecture of FORNAX . The main advantages of the FORNAX code are its efficiency due to its explicit nature (which in itself has sped up the calculation by 3-5 times), its excellent strong scaling to at least ∼ 150,000 cores, its truly multi-dimensional transport solution, and the interior static mesh refinement

We are now at a pivotal time in the theory of core-collapse supernova. Major progress in the past has been in phases. The first phase of the modern era (post-2000) involved the routine simulation in 1D (spherical). This allowed explorations in parameter and progenitor space to fully characterize the phenomenon as a function of all important quantities. Mistakes could be made quickly and an overarching understanding in 1D could be achieved. However, we knew that the cores were unstable to hydrodynamic overturn and turbulence that could not be captured in 1D.

The next phase ( ∼ 2010-2017) was when technique and hardware advanced sufficiently that 2D (axisymmetric) calculations became as routine as 1D had been and many 2D calculations and parameter explorations were enabled. This phase gave us a glimpse of the effects of turbulence and convection with state-of-the-art neutrino transport.

At the end of this phase, a few 3D simulations started appearing, but each of these “heroic” simulations required approximately one year of simulation time on HPC resources. As a result, we obtained only glimpses of the range of outcomes and their dependences on parameters and stellar models in the full 3D of Nature. 

We have now entered what we believe to be the third phase of modern supernova theory, wherein multiple 3D simulations can now be performed each year, each requiring but ∼ one month of wall-clock time. This phase ramped up at the beginning of 2019 and is starting to reveal the full systematic dependence on progenitor, microphysics, and resolution. Fully characterizing the CCSN phenomenon in this way with multiple 3D simulations, with some tolerance for the inevitable mistakes, has been the goal of supernova theory for decades and is the ultimate astrophysics Grand Challenge.