NERSC Initiative for Scientific Exploration (NISE) 2011 Awards

Morphology of Young Core-Collapse Supernova Remnants from Multi-Physics Three-Dimensional Simulations

Tomasz Plewa, Florida State University

Associated NERSC Project: Supernova Explosions in Three Dimensions (m461)

The project will provide data for astrophysics, and highlight current problems in computational hydrodynamics, multiphysics modeling, and high performance computing. It will offer exploratory data for visualization and analytics. Our application results, and the results of our verification and benchmarks, will be fully documented and made publicly available laying a foundation for future collaborations and stimulating the research of our competitors. The results of this project are highly relevant for our high-energy density physics supernova experiment to be executed at the National Ignition Facility.

We propose to expand the research topics of our group into the field of core-collapse supernovae. We have extensive experience in modeling core-collapse supernova explosions, both in modeling the explosion itself (about 1 second after the core bounce) and the early interaction between supernova shock and the progenitor envelope (a few hours after the core bounce). Until now such studies were done using at least two different codes, one for the explosion and another for the later evolution. Furthermore, the explosion phase is almost exclusively modeled in spherical geometry in 3D or by assuming axisymmetry in 2D. This inherently poses a problem at the symmetry axis resulting in significant and hard to quantify discretization errors. This results in having to treat the region near the symmetry axis in a special way, in particular excluding this region from computations. The role of such modifications is very difficult to assess, and it is conceivable they affect the explosion process.  Finally, the central region of the mesh is excluded from models due to time-step restrictions imposed in spherical geometry near the origin (where the neutron star sits). This prevents obtaining self-consistent models that would account for the evolution of the neutron star and its interaction with the post-shock region.

In this research study, we will address some of the above issues. We propose to study the supernova explosion within a single simulation using a single code, starting with the onset of neutrino driven convection until the expansion of stellar ejecta is homologous (several weeks after explosion).  One element required to validate the supernova explosion models is to obtain model observables. For supernovae, observations almost exclusively cover epochs much later (weeks, young supernova remnant phase) than the explosion itself (seconds). This motivates our goal of adapting the hybrid characteristic ray tracing algorithm (HCRT) for photons to neutrino transport in core-collapse supernovae (ccSNe). Such a method also provides a much more realistic treatment of energy deposition, as current non-radiation hydrodynamics (non-RHD) models approximate the effects by assuming an energy deposition distribution. This approach also provides a computationally feasible alternative to Boltzmann neutrino transport models which are on the order of 100 times more expensive. The HCRT method uses parameterized neutrino fluxes as the boundary conditions for the neutrino transport which is expected to yield accurate results as long as the description of neutrino physics does not severely change the ratio of neutrino heating timescale to the timescale of participating hydrodynamic instabilities (e.g. standing accretion shock instability, neutrino driven convection). These simplifications enable parameter studies of supernova explosions in 3D while maintaining a realistic treatment of the physics involved. In addition, implementing HCRT on adaptive Cartesian meshes will allow us to mitigate domain selection choices that have potentially biased previous simulations. By following this path we will also be enabling future studies with complex geometric configurations and neutrino source distributions.

By limiting the assumptions about the progenitor star (in contrast to non-RHD) we will be able to explore possible links between progenitor structure, the explosion mechanism, and young supernova remnant morphology, therefore enabling validation of our models. In particular, we will investigate the SN remnant (SNR) morphology dependence on the rate of stellar core rotation and its structure. This study will supplement the group’s current SN-motivated experimental project at the National Ignition Facility. This project will be the sole focus of one doctoral student with extensive practical experience with our supernova code and the NERSC facility (Timothy Handy).

This implementation will be designed for high performance, distributed memory systems. The HCRT algorithm scales up to 512 cores, and we expect our initial implementation will achieve this level of performance. We will conduct a detailed analysis of parallel performance, identify bottlenecks, improve scalability, and enable computing models with several thousands of processors. We are also interested in exploring speeding up calculations using accelerators (GPUs, FPGAs, etc.) in an effort to prepare for exascale platforms.