3D-Spatiotemporal Forecasting the Expansion of Supernova Shells Using Deep Learning toward High-Resolution Galaxy Simulations
share
Small integration timesteps for a small fraction of short-timescale regions are bottlenecks for high-resolution galaxy simulations using massively parallel computing. This is an urgent issue that needs to be resolved for future higher-resolution galaxy simulations. One possible solution is to use an (approximate) Hamiltonian splitting method, in which only regions requiring small timesteps are integrated with small timesteps, separated from the entire galaxy. In particular, gas affected by supernova (SN) explosions often requires the smallest timestep in such a simulation. To apply the Hamiltonian splitting method to the particles affected by SNe in a smoothed-particle hydrodynamics simulation, we need to identify the regions where such SN-affected particles reside during the subsequent global step (the integration timestep for the entire galaxy) in advance. In this paper, we developed a deep learning model to predict a shell expansion after a SN explosion and an image processing algorithm to identify SN-affected particles in the predicted regions. We found that we can identify more than 95 per cent of the target particles with our method, which is a better identification rate than using an analytic approach with the Sedov-Taylor solution. Combined with the Hamiltonian splitting method, our particle selection method using deep learning will improve the performance of galaxy simulations with extremely high resolution.
READ FULL TEXT